MUCKE aims to mine a large volume of images, to structure them conceptually and to use this conceptual structuring in order to improve large-scale image retrieval. The last decade witnessed important progress concerning low-level image representations. However, there are a number problems which need to be solved in order to unleash the full potential of image mining in applications. The central problem with low-level representations is the mismatch between them and the human interpretation of image content. This problem can be instantiated, for instance, by the incapability of existing descriptors to capture spatial relationships between the concepts represented or by their incapability to convey an explanation of why two images are similar in a content-based image retrieval framework. We start by assessing existing local descriptors for image classification and by proposing to use co-occurrence matrices to better capture spatial relationships in images. The main focus in MUCKE is on cleaning large scale Web image corpora and on proposing image representations which are closer to the human interpretation of images. Consequently, we introduce methods which tackle these two problems and compare results to state of the art methods. Note: some aspects of this deliverable are withheld at this time as they are pending review. Please contact the authors for a preview.

http://ieeexplore.ieee.org/iel2/4524/12820/00592460.pdf

Image Processing

A scheme is developed for classifying the types of motion perceived by a humanlike robot. It is assumed that the robot receives visual images of the scene using a perspective system model. Equations, theorems, concepts, clues, etc., relating the objects, their positions, and their motion to their images on the focal plane are presented. >

http://ieeexplore.ieee.org/iel2/815/3148/00100651.pdf

Robot vision

Many biological applications require the segmentation of cell bodies, membranes and nuclei from microscopy images. Deep learning has enabled great progress on this problem, but current methods are specialized for images that have large training datasets. Here we introduce a generalist, deep learning-based segmentation method called Cellpose, which can precisely segment cells from a wide range of image types and does not require model retraining or parameter adjustments. Cellpose was trained on a new dataset of highly varied images of cells, containing over 70,000 segmented objects. We also demonstrate a three-dimensional (3D) extension of Cellpose that reuses the two-dimensional (2D) model and does not require 3D-labeled data. To support community contributions to the training data, we developed software for manual labeling and for curation of the automated results. Periodically retraining the model on the community-contributed data will ensure that Cellpose improves constantly.

Cellpose: a generalist algorithm for cellular segmentation

In recent years, the deluge of complicated molecular and cellular microscopic images creates compelling challenges for the image computing community. There has been an increasing focus on developing novel image processing, data mining, database and visualization techniques to extract, compare, search and manage the biological knowledge in these data-intensive problems. This emerging new area of bioinformatics can be called ‘bioimage informatics’. This article reviews the advances of this field from several aspects, including applications, key techniques, available tools and resources. Application examples such as high-throughput/high-content phenotyping and atlas building for model organisms demonstrate the importance of bioimage informatics. The essential techniques to the success of these applications, such as bioimage feature identification, segmentation and tracking, registration, annotation, mining, image data management and visualization, are further summarized, along with a brief overview of the available bioimage databases, analysis tools and other resources.

Contact: pengh@janelia.hhmi.org

Supplementary information: Supplementary data are available at Bioinformatics online.

Bioimage informatics

The visual inspection of bridges demands long inspection time and also makes it difficult to access all areas of the bridge. This paper presents a visual-based crack detection technique for the automatic inspection of bridges. The technique collects images from an aerial camera to identify the presence of damage to the structure. The images are captured without controlling angles or positioning of cameras so there is no need for calibration. This allows the extracting of images of damage sensitive areas from different angles to increase detection of damage and decrease false-positive errors. The images can detect cracks regardless of the size or the possibility of not being visible. The effectiveness of this technique can be used to successfully detect cracks near bolts.

Vision‐Based Automated Crack Detection for Bridge Inspection

"A picture is worth one thousand words". This proverb comes from Confucius a Chinese philosopher before about 2500 years ago. Now, the essence of these words is universally understood. A picture can be magical in its ability to quickly communicate a complex story or a set of ideas that can be recalled by the viewer later in time. Visual information plays an important role in our society, it will play an increasingly pervasive role in our lives, and there will be a growing need to have these sources processed further. The pictures or images are used in many application areas like architectural and engineering design, fashion, journalism, advertising, entertainment, etc. Thus it provides the necessary opportunity for us to use the abundance of images. However, the knowledge will be useless if one can't _nd it. In the face of the substantive and increasing apace images, how to search and to retrieve the images that we interested with facility is a fatal problem: it brings a necessity for image retrieval systems. As we know, visual features of the images provide a description of their content. Content-based image retrieval (CBIR), emerged as a promising mean for retrieving images and browsing large images databases. CBIR has been a topic of intensive research in recent years. It is the process of retrieving images from a collection based on automatically extracted features.

/pdf/a-survey-of-shape-feature-extraction-techniques-3qmkhziio7.pdf

A Survey of Shape Feature Extraction Techniques

In this study, we present a one dimensional descriptor for the two dimensional object silhouettes which in theory remains absolutely invariant under affine transforms. The proposed descriptor operates on the affine enclosed area. We design a normalizing contour method. After this normalization, the number of points on a contour between two appointed positions doesn't change with affine transforms. We prove that for the filtered contour, the area of a triangle whose vertices are the centroid of the contour and a pair of successive points on the normalized contour is linear under affine transforms. Experimental results indicate that the proposed method is invariant to: boundary starting point variation, affine transforms even in the case of high deformations and noise on shapes in a given. We also propose a method to simulate the noise contaminating the test shapes and define the signal-to-noise ratio for a shape. In addition, the proposed normalization method can be associated to other algorithms for increasing their robustness to affine transforms and decreasing their complexity in similarity measurements. Index Terms—Affine-invariant, Contour, Descriptor, Filter, Noise.

http://www.iaeng.org/IJAM/issues_v36/issue_2/IJAM_36_2_5.pdf

Affine Invariance Contour Descriptor Based on the Equal Area Normalization

In this study, we present a one dimensional descriptor for the two dimensional object silhouettes which in theory remains absolutely in-variant under affine transforms. The proposed descriptor operates on the affine enclosed area. We design a normalizing contour method. After this normalization, the number of points on a contour between two appointed positions doesn't change with affine transforms. We prove that for the fil-tered contour, the area of a triangle whose vertices are the centroid of the contour and a pair of successive points on the normalized contour is linear under affine transforms. Experimental results indicate that the proposed method is invariant to: boundary starting point variation, affine transforms even in the case of high deformations and serious noise on shapes. We also propose a method to simulate the noise contaminating the test shapes and define the signal-to-noise ratio for a shape. In addition, the proposed nor-malization method can be associated to other algorithms for increasing their robustness to affine transforms and decreasing their complexity in similarity measurements.

Contour Descriptor Based on Affine Invariance Normalization

In this paper the relative-chord context is proposed for shape description in object recognition. The relative-chord is introduced to obtain the affine-invariant feature. The intervals of parallel chords are normalized by vertical direction-chord. Experimental result based on the Columbia University Coil-100 3D database demonstrates the feasibility of the relative-chord context methodology and also highlights its invariance to affine transforms.

Ronsin Joseph

Papers

A Survey of Shape Feature Extraction Techniques

Affine Invariance Contour Descriptor Based on the Equal Area Normalization

Contour Descriptor Based on Affine Invariance Normalization

Object recognition using relative-chord context