An efficient texture-structure conserving patch matching algorithm for inpainting mural images

Preserving heritage paintings across the globe has nowadays gained momentum to let know artistic values of our ancestors in terms of their art techniques and natural material used in creating variety of magnificent paintings so that remains as witness and evidences of ancient historical and cultural heritage. Also, reconstruction of degraded medical images for proper diagnosis is crucial concern for the medical industry. An efficient texture-structure conserving patch matching algorithm (TSCPMA) has been proposed to inpaint the degraded region of an image. The novel feature of Criminisi algorithm to generate large missing areas and reconstruct small gaps is enhanced by improving quality of inpainting and removing existing drawbacks. The priority dependency on confidence and data had been removed by selecting patch to reconstruct with least number of unknown elements. The criteria for minimum similarity distance to select the best patch match had been refined for better patch match thus improving inpainting quality. The target pixel is assigned final value after all unknown pixels from the degraded region have been estimated. The look up table is updated at each iteration so that neighbourhood information can better relate adjacent pixels rather than approximating them with values from other distant known regions of the image. The proposed TSCPMA is able to preserve the color, textural and structural quality of the reconstructed patches as indicated by inpainted results and performance parameters when compared with state of art methods.