Education,
(its still the most Important.....)

Ph.D. Computer Science, Tel-Aviv University, Israel .
Research interests: Computer Vision, Image Processing and Signal Understanding.
Concentrations: Graph theory models in adaptive linear segmentation for video processing.
Dissertation: Advanced Spatial and Temporal Segmentation models and Their Applications.M.Sc. Computer Science , cum-laude, Tel-Aviv University, Israel.
Research interests: Image Processing (models of Illumination in still images)
Thesis: Illumination independent change detection algorithm in a pair of gray images based on connectivity analysis along gray levelsB.A. Computer Science, Tel-Aviv Academic College, Israel .
Concentrations: Graph theory models.

***It important to mentation here my deepest gratitude to my doctoral supervisor , Professor Amir Averbuch . I am profoundly thankful to Amir, whose enthusiasm, patience, wisdom, and unwavering dedication to his students proved invaluable—particularly during moments when the task at hand appeared daunting. The academic freedom Amir granted for the exploration of new ideas and approaches, while simultaneously guiding me in the right direction, rendered the four years of research under his mentorship both engaging and intellectually stimulating.

Videos from the research at TLV University

Illumination independent Object based Change Detection Article

Graph based Segmentation of Moving Objects while spatial and temporal information are available.

Color Image Segmentation Based on Adaptive Local Thresholds

Tracking of Moving Objects Based on Matching between Graph Edges

Some videos of the past work for Artimedia

• Roll up

Foundations of Computer Vision

Reminder : we are creating the AI using our accumulated knowledge over the history, the AI just uses the knowledge better and faster ......

• Segmentation
• Change Detect
• Tracking Obj
• C Vision via AI

• Main Page

NOTE: In much of my research, I rely on graph-based data structures and their associated algorithms including : breadth-first-search (BFS), depth-first-search (DFS), graph contraction, minimum spanning tree (MST), finding k shortest path between two vertices, to obtain efficient implementations. Thus, provides a set of related constructive algorithms for high-level processing that has linear, almost-linear and polynomial time complexity. Linear or almost linear time-complexity algorithms are proportional to the image size n. Thus , Its enables getting a polynomial time complexity for some algorithms when the complexity is proportional to number of arcs E in the image segmentation boundaries. Then the complexity becomes O(E x E) rather than O(N x N) and E<<N.

Segmentation

Is it possible to segment "correctly" with no prior information?

Color Image Segmentation Based on Adaptive Local Thresholds !!!!

An algorithm that integrates edges and region-based techniques while local information is considered. The local consideration enables to derive local thresholds adaptively such that any threshold is associated with a specific region represented by graph. The number of thresholds is automatically determined during the process, which is also automatically terminated

Here is the official publication Link of the above article.

What's happens when we have both saptio and temporal information

Graph based Segmentation of Moving Objects Based on Connectivity Analysis of Spatio-temporal information

Lets dive into a Segmentation algorithm for separating moving objects from the background in video sequences without any prior information of their nature !. I formulate the problem as a connectivity analysis of region adjacency graph (RAG) based on temporal information. By performing a watershed based algorithm I managed to segment the frame into a semantic homogeneous region. The boundary pixels in each region are compared with a series of consecutive frames in order to generate temporal information. The edges of the RAG will represents the temporal information. Each node represents a different homogeneous region. Analysis of the RAG's connectivity is achieved by performing a modification of the breadth-first-search (BFS) algorithm. After a sufficient number of comparisons each of the object's components is merged into a single segment which represents the moving object in the frame. The accuracy of the algorithm is proportional to the number of allowed comparisons.

Here is the official publication Link of the above article.

• Roll Up

Change Detection

Identify changes between two gray images taken in extreme! different illumination, is it possible ?

Illumination independent change detection technique between two gray images, it possible !

The goal of such algorithm is to extract the objects that appear only in one of two registered images. A typical application is surveillance, where a scene is sampled at different time gap. Assumption of significant illumination difference between the two images is considered. For example, one image may be captured during daylight while the other image may be captured at night with infrared device. Now the secret for the solution here is by analyzing the connectivity along gray-levels, all the blobs that are candidates to be classified as ‘change’ are extracted from both images. Then, the candidate blobs from both images are analyzed. A Blob from one image that has no matched blob in the other image is considered as a ‘change’. The algorithm is reliable, fast, accurate, and robust even under significant changes in illumination. The worst-case time complexity of the algorithm is almost linear in the image size. Therefore, it is suitable for real-time applications.

Here is the publication Link of the above article.

So The above algorithms are also was considered to be the Most Cited paper !!!

• Roll Up

Smart Tracking

Can we track after occluded objects in linear complexity ?

Yes !! , but for linear complexity, a graph based theory algorithm is required

The above Algorithm suggests a novel algorithm for tracking moving objects in video sequences. The proposed algorithm is contour-based. The region adjacency graph (RAG) that is obtained from the segmentation results of the input frame, is used to segment the object's contour into subcurves. Then, each connection between two subcurves is called an 'important' junction. The motion of the junctions is estimated in a search that is represented by the RAG edges in the consecutive frame. Each pair of matched junctions may be connected by several edge paths that are candidates to represent the tracked contour. The candidate paths are obtained by an algorithm that finds the k shortest paths between two nodes. It operates on an RAG that is transformed into a weighted and directed graph. Finally, a construction of the tracked contour is achieved by a match process between the object edges and the set of candidate paths.

Here is the publication Link of the above article.

• Roll Up

AI & Computer Vision

If we understand how AI was really built, we can leverage its intelligent way way further than we can "think" !

Lets start with understanding of what is The K-means algorithm: so its an unsupervised machine learning algorithm used to group data points into k clusters based on their similarity, with each data point assigned to the cluster with the nearest centroid (cluster center). It works by iteratively assigning data points to clusters and then updating the centroids to the mean of the assigned points, repeating the process until no data points change cluster membership. K-means is useful for applications like customer segmentation and image analysis but requires the number of clusters (k) to be specified beforehand and is best at finding spherical clusters.

Wait , it still under construction soon to be continue............

• Roll up