Ashfaqur Rahman

Feature weighting methods for abstract features applicable to motion based video indexing

Content based labels, associated with image sequences in contemporary video indexing methods, can be textual, numerical as well as abstract, including colour-histograms and motion co-occurrence matrices. Abstract features or indices are not explicitly numeric entities but rather are composed of numeric entities. When multiple abstract features are involved, distance metrics between image sequences need to be weighted. Most feature weighting methods in the literature assume that the space is numeric (either discrete or continuous) and so not applicable to abstract feature weighting. This paper elaborates some feature weighting methods applicable to abstract features and both binary (feature selection) and real-valued weighting methods are discussed. The performance of different feature selection and weighting methods are provided and a comparative study based on motion classification-experiments is presented.

A Temporal Texture Characterization Technique Using Block-Based Approximated Motion Measure

Characterized by their distinctive motion patterns, temporal textures are natural phenomenon exhibiting spatio-temporal regularity with indeterminate spatial and temporal extent. This paper presents a real-time motion-based temporal texture characterization technique for the first time using block-based motion measures with very high classification accuracy against the popular opinion that such an accurate characterization is only possible using pixel-based measures. Finding an optimal weight ratio between space and time domain features where the accuracy of this block-based technique peaks has been the essence of this success. Computational complexity analyses and classification results clearly demonstrate the capability of the proposed technique in producing comprehensive classification results comparable to the best pixel-based technique with overwhelming reduction in computational complexity.

Real-time temporal texture characterisation using block-based motion co-occurrence statistics

Contemporary temporal texture classification methods use pixel based features thus making the process slow for time sensitive applications like video indexing and surveillance. In this paper, a real-time classification technique is presented by using readily available block based motion vectors. Experimental results demonstrate the ability of the proposed technique to classify a large set of temporal textures in real-time with high accuracy.

A robust optical flow estimation algorithm for temporal textures

In this paper we propose a robust optical flow estimation algorithm for objects whose motion patterns are classified as temporal textures. While suitability of contemporary motion estimation algorithms depends on physical properties of textures, the proposed technique is robust enough to compute accurate flow patterns for a wide range of temporal textures independent of their type. Experimental results demonstrate the capability of the proposed technique producing accurate flow estimates for a wide range of temporal textures.

Segmentation of dynamic textures

Dynamic textures are textures with motion. Characterization of visual processes consisting of multiple dynamic textures is of vital importance to computer vision research, with a diverse set of applications in the field of robot navigation, and remote monitoring applications etc. In the current literature, however, studies are mostly limited to characterization of single dynamic textures. In this paper we aim to address the problem of segmenting image sequences consisting of multiple dynamic textures. More precisely we separate image segments having different characteristic motion patterns - a key attribute of individual dynamic textures. Experimental results demonstrate the ability of the proposed technique by segmenting a wide variety of multiple dynamic texture image sequences.

A Motion-Based Approach for Temporal Texture Synthesis

In this paper we propose a motion based approach for temporal texture synthesis. Temporal textures are natural phenomenon characterized by their distinctive motion patterns. Synthesis of these textures is thus considered as regeneration of motion pattern that has identical motion distribution of a source texture. In this paper we propose a synthesis technique where new textures are generated by computing their movement pattern from a known motion distribution followed by generation of image frames. Experimental results demonstrate the ability of the proposed technique by producing visually promising temporal textures.

Multiple temporal texture detection using feature space mapping

Characterization of visual processes consisting of multiple temporal textures is of vital importance to computer vision research, with a diverse set of applications in the field of robot navigation, remote monitoring (for the prevention of natural disasters), traffic monitoring, and homeland security applications etc. In the current literature of temporal textures, however studies are mostly limited to characterization of single temporal textures. In this paper we aim to address the problem of detecting the presence of multiple temporal textures in an image sequence for real time situations like bushfire monitoring by establishing a correspondence between feature space of temporal textures and that of their mixture in an image sequence. Accuracy of our proposed technique is both analytically and empirically established with detection experiments yielding 92.5% average accurate detection rate on a diverse set of temporal texture mixtures.

A new video indexing and retrieval method for temporal textures using block-based cooccurrence statistics

A new unified video indexing and retrieval method is presented to classify temporal texture videos using spatial as well as temporal cooccurrence statistics of block-based motion vectors, so keeping the computational complexity for retrieval within a real-time bound. Experimental results clearly demonstrate the superiority of the proposed method over existing temporal cooccurrence matrix-based solutions.

A motion-based approach for segmenting dynamic textures

Characterisation of visual processes consisting of multiple dynamic textures is of vital importance to computer vision research, with a diverse set of applications in the field of robot navigation and remote monitoring applications, etc. In the current literature, however, studies are mostly limited to characterisation of single dynamic textures. In this paper, we aim to address the problem of segmenting image sequences consisting of multiple dynamic textures. More precisely, we separate image segments having different characteristic motion patterns – a key attribute of individual dynamic textures. Experimental results demonstrate the ability of the proposed technique by segmenting a wide variety of multiple dynamic texture image sequences.

Periodicity estimation of Dynamic Textures

Dynamic Textures (DTs) are image sequences of natural events like fire, smoke, water etc., that possesses regular motion patterns. Periodicity is a widely used tool to analyse regular structures of periodic one dimensional signals as well as two dimensional image textures. In this paper we present a technique to compute periodicity of regular motion patterns of DT. The proposed technique is based on co-occurrence matrix calculation – another commonly used tool in image texture analysis. Experimental results demonstrate the ability of the proposed technique to categorise the DT in terms of their periodicity and achieving good classification results using computed periods.