Novel Post-Processing Methods Used in Detection of Blotches in Image Sequences

Abstract

SummaryBlotches are the common artifacts in degraded motion picture sequences. They are usually caused by placing dust and dirt on film surfaces as well as abrasion of film materials. Blotches are seen as dark and bright flashes spreading through the frames randomly. The spike detection index (SDIa) method is the simplest approach used to detect these artifacts. However, the drawback of SDIa method is that when the motion vectors are not precise enough in some points (e.g. in edges of moving objects) these points might be declared as blotches too. This situation can also occur in areas containing a high amount of noise. To overcome these difficulties, two post-processing methods are proposed in this paper. In the first method, the edge points are first omitted from the set of detected points and then to restore true edges, the constrained dilation algorithm is applied. In the second method, we use the local average error value in the detection process. According to low intensity variations in blotch areas, the points with relatively close intensity values are used in the averaging process. Additionally, the combination of SDIa and AR methods are proposed to be used in rather large blotch regions to achieve better detection results. Moreover, to implement the intensity interpolation stage, we propose to use adaptive block sizes to achieve better results. In this paper, a performance comparison among the available methods of detection is stated which clearly shows the superiority of the proposed methods.

Detection of Blotch and Scratch in Video Based on Video Decomposition

“…It usually brings out some false alarms arising from local object motion or incorrect global motion compensation. Morphological postprocessing has been proposed in [4], which obtains higher detection rates than SDIa but false alarms are not eliminated adequately. In [2], a ranked order difference (ROD) detector is designed to arrange pixels from motion compensated previous and subsequent image regions, and adopt a three-stage thresholding to them.…”

Trainable blotch detection on high resolution archive films minimizing the human interaction

“…Blotch detector algorithms can be divided into three main groups: (1) spatial detection by analysis of the contrast or local maxima/minima; (2) spatio-temporal methods based on the detection of temporal discontinuities; (3) hybrid methods which are the mixture of the previous two groups [4,11,12]. The first group includes fast morphological filters [13,14] to detect impulsive distortions but with limited capability to detect large blotches with low contrast.…”

“…In the statistical evaluation of the performance of the detection methods, usually some predefined artificial blotch model [4][5][6][7] is considered for both of the shape and the surface in the generation of the test data. In our case, this approach cannot be applied because the proposed object-based classification method analyzes the image intensity features, which would be effected by the artificial blotch generation model.…”