Computer Vision
Digital Geometry
Computational Theory and Mathematics
Document Image Processing
Low-level Image Processing and Representation
Md Ajij - Awarded, September 2023, Awarded from National Institute of Technology Meghalaya, Shillong.
Thesis: Quasi-straight Edge Segments: An Exploration and Related Applications.
Sourav Samanta - Awarded, June 2025, Awarded from Indian Institute of Information Technology, Kalyani.
Thesis: Empowering Smart Agriculture through Computer Vision-based Crop Disease and Quality Monitoring.
Ongoing (Registered at Indian Institute of Information Technology Kalyani)
Alokeparna Choudhury: Area of research - Enhanced Firefly Algorithm and Image Segmentation.
Shiplu Das: Area of research - Deep Learning Models for Driver Drowsiness Detection.
Atashi Saha: Area of research - Machine Vision and Applications.
Title of the Research Project: Automatic Image-based Crop Disease Detection and Severity Estimation
Duration: 3 Years (Commencing from March, 2024)
Funded by: Department of Science and Technology and Biotechnology, Govt. of West Bengal
PI: Dr. Sanjoy Pratihar
Co-PI: Dr. Sanjay Chatterji
Title of the Research Project: AI for Agriculture & Food Sustainability
Duration: 3 Years (Commencing from March 01, 2020)
Funded by: Ministry of Electronics and Information Technology (MeitY), R&D in IT Group, ITEA Division, Govt.
of India, Electronics Niketan, New Delhi
Investigators: Dr. Sanjay Chatterji, CSE, IIIT Kalyani, Dr. Sanjoy Pratihar, CSE, IIIT Kalyani, Dr. Imon Mukherjee, CSE, IIIT Kalyani
Others
Received International Travel Support to attend the 19th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP: VISAPP 2024), held in Rome, Italy, 2024. Funding Agency: Science and Engineering Research Board (SERB), Dept. of Science & Technology (DST), Govt. of India.
Sanjoy Pratihar, On Farey Sequence and Farey Table with Digital-geometric Applications to Image Analysis, Indian Institute of Technology, Kharagpur, India, 2015. [Thesis-Intro-pdf]
Thesis Advisor: Prof. Partha Bhowmick
The Farey sequence, F_n, of order n is the sequence of simple/irreducible, proper, positive fractions with denominators up to n, arranged in increasing order. The concept is well-known in theory of fractions, but from the algorithmic point of view, very limited work has been done so far. In our work, we have augmented a Farey sequence with compound fractions, improper fractions, and negative fractions, which do not find any place in the original sequence.
With all these fraction ranks, we build the Augmented Farey Table (AFT). Farey sequences up to order five are shown in the left. A computationally interesting problem addressed in recent time is the rank problem and its associated order statistic problem. Given the order n, the rank problem is to find the rank of a given fraction in the Farey sequence, whereas the order statistic problem deals with finding the fraction in Farey sequence for some given rank. We have addressed these problems and have shown how they can be used efficiently in many digital geometric applications. The AFT becomes a natural choice in digital-geometric techniques, since the best-known algorithm can compute the rank of a fraction in no sooner than O(n2/3log(n/3)) time. One of the novel features of an AFT is that computations on the digital plane used by different algorithms can be crafted down to simple integer operations.
We have derived several important characterizations of a Farey sequence in order to design an efficient algorithm for its generation. We have also shown how, for a fraction not present in a Farey sequence, the rank of the nearest fraction in that sequence can efficiently be obtained from the concerned AFT. We have derived some useful properties of AFT. Based on these properties, we propose an alternative representation which is useful when the AFT becomes unmanageably large in size for a large order of Farey sequence. We show how the new representation helps in efficiently solving both the rank problem and the order statistics problem.
Generation of F_10 : [0, 1/2]
Augmented Farey table (AFT) of order 10 from augmented Farey sequence.
AFT (n=4); Augmented Farey sequence (n=4).
Polygonal approximation An efficient boundary representation of an object in the digital plane is done through polygonal approximation. During approximation, reasonably collinear straight edges are successively merged. The collinearity is tested by edge slope, which corresponds to AFT rank. If the rank difference (differential Farey index) of two edges is less than a prescribed tolerance, then the two edges are merged into a single edge in an iterative manner. The AFT has been used by us for polygonal approximation in gray-scale images.
Shape Representation If all the internal angles are written in order for a polygon, we get an idea about its shape. As a novel alternative, we have used the sequence of rank differences (differential Farey indices) corresponding to adjacent edges. This has subsequently been used for shape decomposition, shape matching, etc.
Vectorization of Thick Digital Lines Vectorization of a digital object provides a succinct, space-efficient, and useful representation for several applications in computer graphics and image analysis. As a fast and efficient vectorization of digitized engineering drawings, we have used AFT for geometric analysis and refinement.
Correction of skew An algorithm for detection and correction of skews present in scanned document images is proposed which uses certain periodic properties of digital straightness directly on gray-scale images, in tandem with the ranks of fractions in a Farey sequence.