Showing papers on "Bilateral filter published in 2002"

Fast bilateral filtering for the display of high-dynamic-range images

Abstract: We present a new technique for the display of high-dynamic-range images, which reduces the contrast while preserving detail. It is based on a two-scale decomposition of the image into a base layer,...

Fast bilateral filtering for the display of high-dynamic-range images

Abstract: We present a new technique for the display of high-dynamic-range images, which reduces the contrast while preserving detail. It is based on a two-scale decomposition of the image into a base layer, encoding large-scale variations, and a detail layer. Only the base layer has its contrast reduced, thereby preserving detail. The base layer is obtained using an edge-preserving filter called the bilateral filter. This is a non-linear filter, where the weight of each pixel is computed using a Gaussian in the spatial domain multiplied by an influence function in the intensity domain that decreases the weight of pixels with large intensity differences. We express bilateral filtering in the framework of robust statistics and show how it relates to anisotropic diffusion. We then accelerate bilateral filtering by using a piecewise-linear approximation in the intensity domain and appropriate subsampling. This results in a speed-up of two orders of magnitude. The method is fast and requires no parameter setting.

On the origin of the bilateral filter and ways to improve it

Abstract: Additive noise removal from a given signal is an important problem in signal processing. Among the most appealing aspects of this field are the ability to refer it to a well-established theory, and the fact that the proposed algorithms in this field are efficient and practical. Adaptive methods based on anisotropic diffusion (AD), weighted least squares (WLS), and robust estimation (RE) were proposed as iterative locally adaptive machines for noise removal. Tomasi and Manduchi (see Proc. 6th Int. Conf. Computer Vision, New Delhi, India, p.839-46, 1998) proposed an alternative noniterative bilateral filter for removing noise from images. This filter was shown to give similar and possibly better results to the ones obtained by iterative approaches. However, the bilateral filter was proposed as an intuitive tool without theoretical connection to the classical approaches. We propose such a bridge, and show that the bilateral filter also emerges from the Bayesian approach, as a single iteration of some well-known iterative algorithm. Based on this observation, we also show how the bilateral filter can be improved and extended to treat more general reconstruction problems.

Fundamental relationship between bilateral filtering, adaptive smoothing, and the nonlinear diffusion equation

Abstract: In this paper, the relationship between bilateral filtering and anisotropic diffusion is examined. The bilateral filtering approach represents a large class of nonlinear digital image filters. We first explore the connection between anisotropic diffusion and adaptive smoothing, and then the connection between adaptive smoothing and bilateral filtering. Previously, adaptive smoothing was considered to be an inconsistent approximation to the nonlinear diffusion equation. We extend adaptive smoothing to make it consistent, thus enabling a unified viewpoint that relates nonlinear digital image filters and the nonlinear diffusion equation.

Smoothing low-SNR molecular images via anisotropic median-diffusion

Abstract: We propose a new anisotropic diffusion filter for denoising low-signal-to-noise molecular images. This filter, which incorporates a median filter into the diffusion steps, is called an anisotropic median-diffusion filter. This hybrid filter achieved much better noise suppression with minimum edge blurring compared with the original anisotropic diffusion filter when it was tested on an image created based on a molecular image model. The universal quality index, proposed in this paper to measure the effectiveness of denoising, suggests that the anisotropic median-diffusion filter can retain adherence to the original image intensities and contrasts better than other filters. In addition, the performance of the filter is less sensitive to the selection of the image gradient threshold during diffusion, thus making automatic image denoising easier than with the original anisotropic diffusion filter. The anisotropic median-diffusion filter also achieved good denoising results on a piecewise-smooth natural image and real Raman molecular images.

Method of hidden surface reconstruction for creating accurate three-dimensional images converted from two-dimensional images

Abstract: An exemplary method of reconstructing a hidden surface area in an image includes the steps of: identifying a separation pixel gap in an image resulting from a repositioning of image pixels to create a three-dimensional image; identifying pixels from one or more images; and filling the separation pixel gap with the pixels. Replacement pixels can be retrieved from a single image in a search sequence of images, or from a plurality of images in the sequence. One or more pixel fill techniques can be employed to determine replacement pixels for the separation pixel gap which best match adjacent background pixels.

On the equivalence of local-mode finding, robust estimation and mean-shift analysis as used in early vision tasks

Abstract: In this paper we show the equivalence of three techniques used in image processing: local-mode finding, robust-estimation and mean-shift analysis. The computational common element in all these image operators is the spatial-tonal normalized convolution, an image operator that generalizes the bilateral filter.

Bilateral filtering in a demosaicing process

Abstract: Demosaicing convolution kernels (kp,c) are incorporated into the framework of bilateral filtering in order to reduce artifacts at abrupt intensity transitions when a color mosaic pattern (10) is converted to an output image (20, 22 and 24). As a consequence of the bilateral filtering within a window that is selected without intensity considerations, intensity values of pixels I(a,b) that are physically close are given greater weight than intensity values of more distant pixels and, simultaneously, intensity values that are quantitatively similar (i.e. photometrically similar) are given greater weight than intensity values that are quantitatively dissimilar. Using photometric similarity in a demosaicing operation reduces the effects of pixels on one side of an abrupt intensity transition in determining interpolated intensity values for pixels on the opposite side of the abrupt intensity transition.

Spatio-temporal filter unit and image display apparatus comprising such a spatio-temporal filter unit

Abstract: Noise reduction is an important feature in consumer television. This is realized by spatial, temporal or spatio-temporal filters. Spatial filters require pixels from within one image, while temporal filters require samples from two or more successive images. The spatio-temporal filter unit (100) according to this invention integrates spatial and implicit motion-compensated temporal noise reduction in one filter. For the motion compensation no motion vectors are required. The spatio-temporal filter unit (100) is provided with a sigma filter (112) comprising one filter kernel (107) designed to operate on the pixels from both a current image and from the output of the spatio-temporal filter unit, being a temporally recursive filtered image. The operation of the spatio-temporal filter unit (100) can be adjusted by varying the thresholds of the sigma filter (112) and the selection of pixels. The adjustments can be controlled by a motion estimator (222), a motion detector (224) and a noise estimator (220).

Programmable and adaptive temporal filter for video encoding

Abstract: A technique is provided for programmably and adaptively temporally filtering pixel values of frames of a sequence of video frames. The technique includes determining a pixel value difference between a pixel of a current frame and a corresponding pixel of a temporally previous frame; and adaptively filtering the pixel of the current frame using a selected filter coefficient. The filter coefficient is selected employing the pixel value difference. For example, multiple thresholds could be employed to differentiate between multiple filter coefficients, with the pixel value difference being employed to determine which filter coefficient is selected for the adaptive filtering. The thresholds and the filter coefficients can also be programmable. Further, the temporal filter can be integrated with a repeat field detection unit of a motion video encoder in order to conserve memory bandwidth.

Electronic endoscope apparatus to which electronic endoscopes with different numbers of pixels can be connected

Abstract: With a device that connects a processor device to electronic endoscopes having CCDs with different numbers of pixels, for example, 410, 270, or 190 thousand pixels mounted thereon, all the CCDs are driven at a frequency corresponding to 410 thousand pixels. Further, an information amount converting circuit enlarges an image obtained by the CCD with 270 or 190 thousand pixels in a horizontal direction and a vertical direction by using pixel interpolation. That is, data on a pixel to be interpolated is calculated by averaging pixels around this pixel, and enlarged binary data on this target pixel is obtained. A weighting factor is applied to this binary data, and the resultant data is added to the target pixel data. This process forms an image with an aspect ratio of 4:3.

Method and system for removing artifacts in compressed images

Abstract: In a method for removing artifacts in a digital image that has been decoded from a compressed representation of the image, the pixels of the digital image are segmented into low detail pixels, high detail pixels, and boundary pixels, and then a single non-linear filtering method, such as a sigma filtering method, is applied to each pixel in the digital image. By selecting separate filter parameters for the low detail, high detail, and boundary pixels, the removal of artifacts is improved over prior methods.

Blending multiple images using local and global information

Abstract: A method blends multiple input images into an output image for any arbitrary view. In the output images, pixels that are produced from only a single input pixel are identified. The weight of the single pixels is set to one. For each remaining pixel in the input images with unassigned weights distances to an image and a depth boundary are measured, and proportional weight, in a range from zero to one, for these remaining pixels are set proportional to the minimum of the two distances. Then, each input image is rendered to the output image according to the blending fields.

Digital image appearance enhancement and compressibility improvement method and system

Abstract: A method and system for processing a digital image and improving the appearance of the image while enhancing the compressibility of the image. The digital image has a plurality of input pixels. The image processing system has a filter selection mechanism for receiving a filter selection window corresponding to a current input pixel and responsive thereto for generating a filter identifier based on either an edge parameter computed based on the filter selection window or an activity metric computed based on the filter selection window. A filter application unit that is coupled to the filter selection mechanism for receiving the filter identifier and applying a filter identified by the filter identifier to an input pixel window to generate an output pixel is also provided.

Programmable horizontal filter with noise reduction and image scaling for video encoding system

Abstract: A technique is provided for programmably horizontally filtering pixel values of frames of a plurality of video frames. The technique includes, in one embodiment, passing pixel values through a real-time horizontal filter disposed as preprocessing logic of a video encode system. The horizontal filter is programmable and includes a filter coefficients buffer for holding multiple sets of filter coefficients. The horizontal filter programmably employs the multiple sets of filter coefficients to selectively perform spatial noise filtering, or spatial noise filtering and image scaling on the pixels. The filter coefficients are also programmable and may be changed dynamically and repeatedly, with changes being applied at frame boundaries. When performing image scaling, multiple sets of filter coefficients are employed.

Adaptive post-filtering for reducing noise in highly compressed image/video coding

Abstract: A technique to reduce ringing artifacts in highly compressed block-based image/video coding is applied to each reconstructed frame output from the decoder. For each pixel block of a reconstructed frame, low-pass filtering is then adaptively applied according certain calculated differences between adjacent pixel values. For each pixel, a determination is made as to what type of horizontal filter, if any, is to be applied. Depending on the results of that determination, the pixel may remain unfiltered or may have a 2- or 3-tap horizontal filter applied to it. A similar process is undertaken to determine what type of vertical filter, if any, is to be applied, no filter, a 2-tap or a 3-tap vertical filter.

Discriminating between changes in lighting and movement of objects in a series of images using different methods depending on optically detectable surface characteristics

Abstract: In a motion detection system, filters are applied to a series of digital images in order to determine whether changes in the properties of pixels between the current image and a reference model are due to motion of objects or to changes in lighting. The filter may be based on scaled differences between a current image and a reference model. The type of surface may be determined based on optical properties of the surface and then only the best filter for that type of surface is applied. Alternately, multiple filters may all be applied and the results combined. The processing may be reduced using a background model to determine which pixels are in the background and which are in the foreground and then only applying the filter to the foreground pixels.

System and method of scaling images using adaptive nearest neighbor

Abstract: A method for interpolating a target pixel from a plurality of source pixels in a high contrast image. The method comprises the following steps. A window of the plurality of source pixels is examined and compared with a plurality of predefined conditions for determining if a structure of significance is present within the window. A filter configuration is selected from a plurality of filter configurations in accordance with results of the comparison. The selected filter is applied to the source pixels for interpolating the target pixel. If the structure of significance is detected in the window, the selected filter best preserves the structure.

Imaging device and signal processing method therefor

Abstract: A color filter arrangement (11) is used, in which a plurality of filter units are each made of 2×2 arrangements of red (R), green (G), green (G) and blue (B) color elements. First, signal charges are added up for all pixels belonging to each of a plurality of pixel blocks made of quadratic arrangements of 3×3 of pixels, which are larger than the filter units (2×2 arrangement). Then, compressed color information for each of the pixel blocks is obtained from a result of the addition for each pixel block, taking the 2×2 arrangements of pixel blocks as large filter units.

Method of performing sub-pixel based edge-directed image interpolation

Abstract: A method of generating a value for a missing pixel “x” by determining a “least harmful” local edge direction between pixels, or sub-pixels, on substantially opposing sides of the missing pixel, and interpolating the difference to arrive at a value for pixel “x”. The method involves generating sub-pixel values for locations within neighboring pixels, the sub-pixels may comprise half-pixels, quarter-pixels, three-quarter pixels, and so forth, wherein any fractional pixel quantity may be created. Absolute difference values are calculated between neighboring pixels, or sub-pixel values, to determine a least harmful local edge direction along which a value is generated for pixel “x” by interpolation.

System and method for edge ehnancement of images

Abstract: The invention provides a method for generating an enhanced image, the method including the steps of: (a) receiving a matrix of pixels representative of an image; (b) generating a gradient image representative of a difference between values of adjacent pixels; (c) calculating a center of mass for each pixel of the gradient image in response to gradient intensity values and location values of neighboring pixels; and (d) generating an enhanced image by modifying intensity values of pixels of the matrix of pixels that are located in a vicinity of local centers of mass in response to intensity values of pixels that are further displaced from the local centers of mass. The invention provides a method for generating an enhanced image, the method including the steps of: (a) receiving a matrix of pixels representative of an image; (b) estimating at least one edge of the image; and (c) generating an enhanced image by modifying intensity values of pixels of the matrix of pixels that are located in a vicinity of the at least one estimated edge in response to intensity values of pixels that are further displaced from the at least one edge.

Comparison of square-pixel and hexagonal-pixel resolution in image processing

Abstract: In most image processing applications, data is collected and displayed in square pixels. Hexagonal pixels offer the advantage of greater rotational symmetry, a close packed structure and a nearly circular pixel. We compared the quality of images using square pixels with that of images employing hexagonal pixels. The comparison was done using various images, each considering a different aspect of geometry (i.e., lines at different angles, curves, etc). The square pixel images were constructed using the average of a square area of smaller square pixels. Hexagonal pixel images were constructed using two techniques. The first one was called the "two-template approach", wherein two different templates were used to create a close packed hexagonal image from smaller square pixels. The second approach was called the "six-neighbor approach"; it creates a rectangular template using the six neighbors of a hexagonal pixel. A Euclidean distance measure was used to compare the square pixel and hexagonal pixel images. A brief explanation of the algorithm and the results are provided. Based on our results obtained using the Euclidean distance as a quality measure, we conclude that, contrary to our intuition and their widespread use in nature (retinas and ommatidia), hexagonal pixels do not offer any advantage over conventional square pixels.

Analysis of the bilateral filter

Abstract: Effective methods for image denoising are typically based on iterative and locally adaptive algorithms. Recently, an alternative algorithm called 'bilateral filter' was proposed for the same task. This filter was shown to give similar and possibly better results compared to the ones obtained by the best iterative approaches. In this paper we present a relation between the bilateral filter and the Bayesian methodology. Based on this observation, we show how the bilateral filter can be improved and extended to treat more general signal characteristics.

Anisotropic vector diffusion in image smoothing

Abstract: Anisotropic diffusion has been widely used in image processing for its efficiency of smoothing the noisy images while preserving the sharp edges. In this paper we explore a general version of anisotropic diffusion schemes for vector-valued images, based on the polar-coordinate representation of the vectors. As an example, we apply our method to color images and show its ability of edge-preserving smoothing on vector-valued images.

Video and graphics system with square graphics pixels

Abstract: A video and graphics system provides square graphics pixels to blend images having 640×480 pixels, such as graphics images provided by some set top boxes and intended to be displayed at a 12.27 MHz display sample rate, with images having 704×480 pixels, such as ITU-R 601 compliant images such as NTSC SDTV images, having oblong pixels and displayed at a 13.5 MHz display sample rate. A sample rate converter including a multi-phase-multi-tap filter is used to generate square pixels. The multi-phase-multi-tap filter provides a good balance of sharpness, smoothness, anti-aliasing and reduced ringing. The multi-phase-multi-tap filter can also be used to convert images having 320×480 pixels to images having 704×480 pixels. The multi-tap filter can be used for scan rate conversion of graphics or video images for HDTV or SDTV applications.

Deblocking block-based video data

Abstract: Block-based video data is deblocked by use of a strong, non-linear filter process or a weak, adaptive smoothing filter process. In the non-linear filter pixels on each side of a block boundary are updated. In the adaptive smoothing filter only one pixel on each side is updated. A smoothness check with reference to configurable threshold values is performed and this is used to dynamically select a relevant filter process. Thus, processing capacity is optimised.

System and method for edge enhancement of images

Abstract: A method for generating an enhanced image includes the steps of: (a) receiving a matrix of pixels representative of an image; (b) generating a gradient image representative of a difference between values of adjacent pixels; (c) calculating a center of mass for each pixel of the gradient image in response to gradient intensity values and location values of neighboring pixels; and (d) generating an enhanced image by modifying intensity values of pixels of the matrix of pixels that are located in a vicinity of local centers of mass in response to intensity values of pixels that are further displaced from the local centers of mass.

A system for controlling gray scale

Abstract: A display matrix is provided for forming a composite image from a series of sub-images In general, the display matrix includes a plurality of display elements (1014), each display element including a pixel, and a display circuit electrically connected to the pixel An illumination source (3002) illuminates each of the pixels Logic controls (3016) the intensity of each pixel utilizing a digitally controlled waveform for controlling a duration of time that illumination is allowed to pass through liquid crystal of the display elements (3014) Intensity values for each pixels are stored in memories (3006,3010) that are associated with the individual pixels of a display Each of the pixels is set to a first state A first intensity level is broadcast to the pixels The first intensity level is analyzed (3004) with the intensity value stored in the pixel memories (3006,3010)

Demosaicking as a bilateral filtering process

Abstract: Digital Still Color Cameras sample the visible spectrum using an array of color filters overlaid on a CCD such that each pixel samples only one color band. The resulting mosaic of color samples is processed to produce a high resolution color image such that a value of a color band not sampled at a certain location is estimated from its neighbors. This is often referred to as 'demosaicking.' In this paper, we approach the process of demosaicking as a bilateral filtering process which is a combination of spatial domain filtering and filtering based on similarity measures. Bilateral filtering smooths images while preserving edges by means of nonlinear combinations of neighboring image pixel values. A bilateral filter can enforce similarity metrics (such as squared error or error in the CIELAB space) between neighbors while performing the typical filtering operations. We have implemented a variety of kernel combinations while performing demosaicking. This approach provides us with a means to denoise, sharpen and demosaic the image simultaneously. We thus have the ability to represent demosaicking algorithms as spatial convolutions. The proposed method along with a variety of existing demosaicking strategies are run on synthetic images and real-world images for comparative purposes.