Low bit rate visual communication using binary sketches for deaf sign language communication

Moorthy, Manoranjan (1998) Low bit rate visual communication using binary sketches for deaf sign language communication. Masters thesis, Memorial University of Newfoundland.

[English] PDF - Accepted Version Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission. Download (5MB)

Abstract

Many commercial video conferencing applications are available to transmit color and gray level images along with voice and text data over telephone lines and Internet protocol networks. However, these applications cannot be used for deaf sign language communication because they offer very low temporal resolution of pictures, at less than a frame per second resulting in jerky movement of objects. Effective sign language communication requires a temporal resolution of the order of 8 to 12 frames per second, while the size of the images may be small. -- The research for this thesis was aimed at providing a practical solution to enable deaf sign language communication using moving cartoons. Since sign language relies on hand shape, orientation, position and movement together with facial expressions, it can be adequately conveyed in moving cartoons or binary sketches. The cartoon or binary sketch is a very economical method of portraying a person or a scene, where emphasis is given to the location of object boundaries at the expense of shading and texture information. An efficient feature extraction algorithm is used to extract cartoon points. This algorithm is based on the postulate that perceptually significant features of the human face and hands, at which cartoon lines should be drawn in the image, occur wherever surfaces in object space are approximately tangential to the line of sight of the camera or viewer. Based on the above postulate, the detector exhibits primary sensitivity to valleys and secondary sensitivity to edges in the image. Adaptive techniques based on the global statistics of the image are proposed to improve the functionality of the feature extractor. To improve the subjective quality of the binary images, irreversible preprocessing techniques, such as threshold hysteresis, isolated point removal and predictive filter are used and results are provided to compare their relative performances. -- A simple and efficient recursive temporal filtering scheme is used to prefilter noisy image sequences from low cost cameras. This scheme uses the spatial homogeneity of the low frequency additive and multiplicative noise, which produces a throbbing or pulsing effect on the image sequences. Histograms of successive frames are used to segment moving and non-moving regions and to formulate a transfer function. The transfer function is used in the filtering operation. Different temporal and spatio-temporal filtering methods are implemented. Subjective and quantitative test results are presented to demonstrate their relative performances. -- An efficient three-dimensional binary image compression scheme is used for binary sketches. This method uses conditional replenishment in the temporal dimension and quad tree or hierarchical coding in the two dimensional space. A complete system is developed for deaf sign language communication over low bandwidth telephone lines and Internet protocol networks. Subjective tests are performed on the system and the results confirm that the system can be used for sign language communication over telephone lines.

Actions (login required)

View Item