Showing papers on "Sum of absolute differences published in 1995"

Fractional pixel motion estimation of video signals

Abstract: Fractional pixel motion estimation of video signals is performed by comparing a block of pixels from a current image of the video signal with a plurality of displaced blocks within a search window from a previous/future reference image of the video signal via a distortion function. The displaced block from the reference image that produces the minimum value for the distortion function provides a center pixel. A general surface is fitted around this center pixel so that it equals the distortion function at each integer pixel location surrounding the center pixel. The distortion function for the fitted surface is estimated for fractional pixel locations, and the motion vector corresponding to the minimum value for the fractional distortion function is selected for transmission as part of the compressed video.

Video encoder and encoding method using intercomparisons of pixel values in selection of appropriation quantization values to yield an amount of encoded data substantialy equal to nominal amount

Abstract: When a digital video signal is compressed, the amount of resulting data is predicted and the quantization level is controlled so as to obtain a constant bit rate. The prediction is made by computing a linear combination of a standard deviation and a number of non-zero coefficients, or by computing a sum of absolute differences between adjacent pixel values, or by computing a dynamic range of pixel values. The bit rate can also be controlled by deleting high-frequency coefficients. To avoid image degradation, the quantization level can also be controlled according to the sensitivity of small image areas to quantization noise. Sensitivity is determined by dividing an area into subblocks, which may overlap, and calculating statistics in each subblock. To reduce the amount of computation required in motion estimation, chrominance motion vectors are derived from luminance motion vectors.

High efficiency encoding apparatus

Abstract: An encoding apparatus which structures a band-divided digital image signal into three-dimensional blocks each having a plurality of pixels, discriminates each three-dimensional block between a moving image block and a static image block, and, in the case of a static image block, encodes only the three-dimensional blocks in the lower frequency bands while dropping the three-dimensional blocks in the higher frequency bands. However, of moving image blocks, blocks having a large pixel variance value are encoded as effective image blocks. Also, different size weightings are performed to transform coefficients depending on whether the block is discriminated as a moving image block or a static image block. Furthermore, shuffling is performed on a group of moving image blocks and a group of static image blocks independently of each other, and the transform coefficients for the static image blocks are quantized after quantizing all transform coefficients for the moving image blocks.

Motion estimation processor architecture for full search block matching

Abstract: A circuit for calculating a sum of absolute errors for use in full block search matching in a motion estimation processor is disclosed herein, the circuit being easily implemented and capable of running at 54 Mhz. The circuit accesses search window and reference data from memory and loads the data into rows of laterally interacting processing elements having an architecture capable of fast data processing. A sum of absolute errors between all elements of each row of search data and all elements of each row of reference data is calculated, and the absolute error for all rows of processing elements is totalled. From this total sum of absolute error, the motion vector may be predicted for the next frame.

Differential motion detection method using background image

Abstract: A differential/absolute value image is derived by calculating an absolute value of each pixel in an image produced by subtracting a background image from an input image. A mean value and a variance of a pixel in the differential/absolute value image are calculated from the levels of pixels included in a small region on the differential/absolute value image which has the pixel located at the center thereof. The calculated mean value and variance of the pixel are used to distinguish whether the pixel is a pixel included in a moving object existing region, a background region, or an input image sudden change region. A method of updating the background image is changed in accordance with whether the pixel is a pixel included in the moving object existing region, the background region, or the input image sudden change region.

Method and system for compressing a pixel map signal using block overlap

Abstract: A pixel map signal is separated into block signals, each representing pixels in a pixel map block (420) in a pixel map (410). The blocks signals each contain an overlap component having pixel values representing pixels contained both by the corresponding pixel map block (420) and an adjacent block. The pixel map signal is compressed by converting the block signals into a coefficient signal of block coefficient signals, each of which represents a pixel map block (420) with the coefficients in a hybrid polynomial. When the pixel map signal is decompressed, the pixel values in the overlap component for each pixel map block (420) are averaged with the pixel values in the overlap component for the adjacent block and replaced by the averaged pixel values in the corresponding pixel map block (420) and the adjacent pixel map block (420).

Image decorative processing apparatus

Abstract: In an image processing apparatus, a matrix for an arbitrary pixel (to be an observed pixel) on the original image is set up so that the density of the observed pixel and that of a target pixel which is located in the direction along which shadows are formed are compared to determine a maximum value of density of the two. If the density of the observed pixel is greater than that of the target pixel or pixels, the density of the target pixel is converted into a predetermined level of density, whereas if the density of the observed pixel is less than that of the target pixel, no conversion of density is made. This procedure is repeatedly performed for the entire image information to thereby create shadows having a designated length and complete a shadowing image. Then, the shadowing image is combined with the original image, whereby a shadowed image is complete.

A segmentation criterion for digital image compression

Abstract: This paper is concerned with segmenting light intensity images for the sake of compressing them using lossy compression techniques. Among the most commonly used techniques for image segmentation is quad-tree partitioning. In this technique, block variance based criteria are usually used to measure the smoothness of the segmented blocks and to consequently classify them. Block variance, however, does not consider the pixel value distribution within the block. Instead of using the block variance as a segmentation and classification measure, we propose using the mean squared deviation from the neighboring pixels mean. The proposed measure is capable of differentiating between blocks not only according to block pixel values but also according to their distribution within the block. This leads to a much better image segmentation and consequently to higher image compression ratios with lower image degradation. The results show the superiority of the proposed measure over the block variance measure.

Image processing method utilizing error diffusion technique

Abstract: The present invention provides a half-tone image processing method utilizing an error diffusion technique which allows for the density adjustment of a half-tone image to be reproduced. When the density of an object pixel is converted into binary-coded data, a binary-coding object value is first calculated by adding a density value of the object pixel to an error sum of binary-coding errors distributed to the object pixel from peripheral pixels around the object pixel. Then, the binary-coding object value is compared with a threshold value TH for judging whether the object pixel is a black pixel or a white pixel. The binary-coding object value is also compared with a reference value GSLVB or GSLVW which can be variably set for the calculation of a binary-coding error HG of the object pixel. The level of the binary-coding error HG can be adjusted by variably setting the reference values GSLVB and GSLVW. Thus, the density adjustment of a half-tone image can be realized.

Reduced bit number motion vector detection.

Abstract: In image compression using motion compensated predictive coding, the number of bits of pixel data is reduced before the pixel data are supplied to a motion vector detecting apparatus. Motion vectors are detected using a pipelined operation in which pixel by pixel differences between a reference block and a search block are obtained and summed. At each stage of the pipeline, the sum of differences is limited to a predetermined maximum number of bits.

Improved post-processing method for use in an image signal decoding system

Abstract: A post-processing method for use in an image signal decoding system, capable of improving the image quality of the system, comprises the steps of: filtering target pixel data stored in a memory; calculating an absolute difference value between an original target pixel data and the filtered target pixel data; updating the stored target pixel data with the filtered target pixel data if the absolute difference value is smaller than a predetermined threshold value; repeating said steps of the filtering to the updating N times as long as the absolute difference value is smaller than the predetermined threshold value and updating the stored target pixel value with compensated target pixel data if the absolute difference value is equal to or larger than the predetermined threshold value, wherein the compensated target pixel data is provided by adding the original target pixel data and the predetermined threshold value if the original target pixel data is smaller than the filtered target pixel data and by subtracting the predetermined threshold value from the original target pixel data if the original target pixel data is greater than the filtered target pixel data; and repeating said steps of the filtering to the repeating for a next target pixel until all the pixels in the current frame are post-processed.

Image processing method and image processing apparatus for executing gradation conversion of a pixel

Abstract: An image processing method for converting the pixel value of a pixel of interest in a displayed image, wherein pixel values of the pixel of interest and pixels surrounding the same are compared with a predetermined threshold value while a conversion characteristic most suitable for the pixel of interest is selected on the basis of the result of comparison indicating the magnitude relationship, and the pixel value of the pixel of interest is changed according to the most suitable conversion characteristic that is selected. The threshold may have different values for different pixels, may have the same value over one scan line, or may be constant for all pixels.

Method for determinig a destination pixel location from an arbitrary source pixel location during scaling of a bit map image

Abstract: A source image is scaled to a destination image using a method which determines a first destination pixel position corresponding to a given source image pixel. The method comprises the steps of: compiling a scale table having one entry per source pixel in a "pixel group", the end destination pixel of a pixel group positioned where a first accumulation (i.e., "Scale Source Sum") of sets of m pixels, equals a second accumulation (i.e., "Scale Destination Sum") of sets n pixels, the Scale Source Sum derived by successively adding an m pixel set value for each successively read source pixel, and the Scale Destination Sum derived by successively adding an n pixel set value for each written destination pixel; factoring both an n source pixel set value and an m destination pixel set value by y, where y is a largest common denominator of both n and m, to derive "n (factored source)" and "m (factored destination)"; using the n(factored source) to determine an integer number of pixel groups between the first source image pixel and the initial source pixel, and determining a total number of destination pixels encompassed by the integer number of pixel groups; determining a position of the given source image pixel within a pixel group on the source raster image scan line; using the scale table to determine a corresponding destination pixel within the pixel group; finding the first destination pixel and positioning the scaled version of the source image thereat.

Automatic identification of faulty points in digital image

Abstract: The faulty points result from faults in a digitiser generating a digital image from a document image. The digital image contains numerous pixel points, each with a pixel colour value. Numerous documents are scanned for generating several digital images. A pixel colour value is allocated to each pixel point in each image with a pixel characterised by its position and colour. A numerical value of documents is accumulated for each pixel point in an interesting section, in which the subjective pixel point contains a pixel of a common colour value. Finally pixel points are added, which have a count value higher than a threshold value of numerous points.

Video data compression system with comparison to reference pixels

Abstract: The data compression system uses comparison information in a time-corresponding pixel in a coding process. Two context statements are used, one when a pixel is the same as the corresponding reference pixel, and one statement for the remainder code for a pixel, which is dissimilar to its reference.The use of the similarity bits saves processing time, where by coding of the similarity bits the decompressor determines that the pixel is the same as the pixel in the previous image, so that no further processing for this pixel is necessary.