3-D shape recovery from image focus using no-reference sharpness metric based on inherent sharpness

Abstract

Recovering an accurate depth map from its corresponding 2-D images using shape from focus architecture is a convoluted issue in computer vision and signal processing society. This paper contributes a new robust focus measure for 3-D shape recovery based on discrete wavelet transform and inherent sharpness approach. This novel focus measure technique utilizes no-reference sharpness metric based on inherent sharpness approach. The no-reference sharpness metric estimates a perceptual sharpness score based on the coefficients of discrete wavelet transform. To obtain the data of high frequency elements in an image the perceptual sharpness metric utilizes diagonal coefficients and approximated sub-signal of wavelet decomposition. The efficiency of the proposed scheme is evaluated by comparing it with state of art shape from focus approaches by conducting experiments on real and synthetic image sequences. Two global statistical metrics are utilized for performance evaluation by conducting experiments on real world images and synthetic image sequences. The evaluation is estimated on the basis of monotonicity and unimodality of the focus measure curve. The experimented results are then discussed in various forms to support the proposed scheme.