Disparity Image Inpainting with Improved Method Based on Exemplar

Abstract

To combat the problems of disparity image inpainting, an improved exemplar-based image inpainting method is proposed. The original stereo image we select which corresponds to disparity image is decomposed by TV-based decomposition and the structural information is obtained. The patch priority in the position corresponding to the disparity image is calculated by using the structural information, which reflects true characteristics of the original stereo image and disparity image. Experimental results show that inpainting performance is better with the guidance of this patch priority for disparity image.

Research of Enhanced Speech Signal Based on MATLAB Simulation Platform

“…According to the voiceless, voiced signal, wavelet transform coefficients [9][10] in different scales varied, the voicing judgment can be made as:if the input speech segment meets the requirements of: (1) the highest signal energy on the smallest scale (ie, the high-band signal energy being most ); (2) the ratio of the largest scales on the signal energy is less than 0.9 to the minimum scale of the signal energy., which can be judged as the unvoiced speech, otherwise, judged as the voiced one.…”