IMAGE PROCESSING METHODS

Image Processing Methods Used In Motion Pictures

Image Processing Methods Used In Motion Pictures

Introduction

Old motion picture collections represent artistic and historic treasures. Therefore, it is desired to copy these pictures onto new media with acceptable cost and quality. The quality of the copy must be as close as possible to those of the today's standards. Projection process can cause films to be damaged and degraded. Also the physical nature of film material is very susceptible to degradation. Therefore, most of them are already in bad shape or continue to deteriorate(Astola et. al. 2010: 678-689). The typical artifacts in degraded motion picture material are flicker, line-jitter, fading, discoloring, vertical scratches and local random defects. In this work, we propose a method for restoration of local random defects because they mainly appear on scenes. The local random defected areas are caused either by the accumulation of dirt or by the film material being abraded. Therefore, they appear as bright and dark blotches on the scenes, referred to as ''dirt and sparkle'' in motion picture industry(Armstrong et. al. 2008: 83-88). The blotches in noise-corrupted image sequences exhibit temporal discontinuity characteristic and they can be considered as multiple pixel-sized impulsive distortion. The problem of restoration of degraded motion pictures has received considerable attention in the last decades. Before introducing our method, we will briefly review other methods that are close to ours. The successful treatment of artifacts in image sequences mainly involves three processes: motion compensation of the moving objects, detection of the missing regions and interpolation of the detected missing regions (Alparone, 2010: 1462-1467). Most conventional image sequence processing algorithms perform some form of global filtering of frames. The common problem of the spatiotemporal 3D filtering-based restoration methods is to accurately estimate motion in a noisy image sequence. The badly motion-compensated pixels will also appear as temporal discontinuities. Therefore, robust motion estimation techniques with respect to additive and replacement noise must be used. Some multiresolution schemes and bi-directional schemes are widely used as a robust motion estimator in many applications (Armstrong 2009: 45-51). Identification of the missing data regions allows efficient algorithms to be developed to interpolate these missing regions. Defects as dirt and sparkle are usually determined by blotch detection methods, which use temporal discontinuities in image sequences. The first detector involves some heuristics that are discussed by Storey. To detect a blotch, the detector utilizes a thresholding procedure based on the forward and backward nonmotion- compensated frame differences. However, in the presence of large motion, this detector cannot correctly separate moving regions from blotched areas.

Some of the existing methods for detection of defected regions in films are based on spike detection index (SDI) , Markov random field (MRF), 3D auto-regressive (AR) models, and rank ordered differences (ROD)

Due to poor estimation of model coefficients in real situations, the methods based on AR approaches result in poor performance. They would miss blotches with a low intensity difference with preceding and next ...