Research On Image Texture Feature Extraction And Haptic Display

Haptic display, as a kind of human-computer interaction technology, simulates human's interaction with environment or objects in virtual scenes, based on human's haptic perception and manipulation mechanism in real environment, to provide natural and intuitive feelings for operators. Introducing texture into the haptic display process will improve the fidelity of haptic rendering, strengthen operators'perceptual ability of virtual environment and recognition ability of virtual objects. Texture feature extraction and haptic modeling based on image is simple and effective, without strict requirements of hardware and prior information, so it has quite widely application value. This paper conducts study on image texture feature extraction and haptic display.In the stage of image preprocessing, an improved image segmentation method is proposed. This method combines fractional order differentiation and gray level co-occurrence matrix to extract texture feature vector, then uses Markov random field model to segment image. Fuzzy entropy criterion is used after the segmentation to optimize the results. Adding backward fractional integral to forward integral, an image de-noising filter operator is deduced in domain space, which can maximize the preservation of edges and texture while guarantying the noise reduction performance.Different feature extraction frameworks are used for texture images and scenery images. As for texture images, a new differential operation is derived to replace traditional integer-order differential operation by subtracting backward fractional differential from forward fractional differential. It is used to obtain the texture extraction operator on the basis of the Sobel operator. The feature vector of texture in different frequency is generated and then used in the haptic display system. As for scenery images, an image structure/texture decomposition method via relative total variation is firstly used. The output texture image is enhanced by thresholding and power transformation. An flow-based cascade Gaussian filter framework is applied to the output structure image to extract contours. The output is coherent and directly results in a good-quality line illustration, thus can reflect the structure feature. The texture and structure features are integrated and used in the haptic display system.Finally, haptic display software platform is built, together with the haptic device, is formed into the haptic display system.2D-texture mapping is realized, using original image, to render the appearance. And normal map is created, using the output of the above-mentioned feature extraction frameworks, to realize 3D-bump mapping. Then "finger-proxy" algorithm is used to haptically render the virtual mesh and provide the user haptic feedback. Based on this system, experiments were conducted, the results of which verified the superiority of the feature extraction frameworks applied in haptic display system.