A local thresholding method is proposed for the removal of marginal noise in gray-scale document images, whereas a region growing method is devised for binary document images. After the detection of marginal noise regions, different removal methods are performed. Marginal noise detection will reduce an original document image into a smaller image, and then find marginal noise regions according to the shape length and location of the split blocks. The proposed approach consists of two steps which are marginal noise detection and marginal noise deletion. This paper proposes a novel approach to remove marginal noise. The overlapping of marginal noise with meaningful objects makes it difficult to perform the task of segmentation and recognition of document objects. Marginal noise might cover meaningful document objects, such as text, graphics and forms. It usually appears in the front of a large and dark region around the margin of document images. Marginal noise is a common phenomenon in document analysis which results from the scanning of thick documents or skew documents. Image defect models and their associated generators permit a new kind of image database which is explicitly parameterized and indefinitely extensible, alleviating some drawbacks of existing databases. Applications of the generator are described, including a polyfont classifier for ASCII and a single-font classifier for a large alphabet (Tibetan U-Chen), both of which which were constructed with a minimum of manual effort. A pseudo-random image generator implementing the model has been built. Work-in-progress towards a parameterized model of local imaging defects is described, together with a variety of motivating theoretical arguments and empirical evidence. Establishing the essential properties of such models, such as completeness (expressive power) and calibration (closeness of fit to actual image populations) remain open research problems. A lack of explicit quantitative models of imaging defects due to printing, optics, and digitization has retarded progress in some areas of document image analysis, including syntactic and structural approaches.
0 kommentar(er)
