Bit depth: Bit depth is a description of the range of values in gray scale that can be possibly assigned to one pixel. For example, a bit depth of 8 bits/pixel, would enable a single pixel to take 256(2³) values from 0 to 255 on the grayscale (Fig 2). The more the bit depth, greater is the color information stored in the image and consequently higher is the image quality. Thresholding: This is the key to raster output. Thresholding is the process by which the scanner assigns a value to a pixel as a representation of the intensity and color of the image. In color scanning the output is a RGB image, which is a color model in which a three layered pixel matrices are formed, one each for R, G, and B channels Colour topographical maps The specific problem of the interpretation of the contour lines out of a scanned map is difficult to resolve not just because of the nature of the raster data set, but because of global constraint due to their topology. Previous studies have shown that the local geometry of the contour lines does not provide evidence for an automatic reconstruction of the relief. To overcome this most of the research works concentrate involving the intervention of an operator. The role of the human being is to generally resolve the ambiguities and/or to correct the result of the reconstruction. Research in image processing, computational geometry, image representation and pattern recognition has let to the development of wide variety of algorithm for contour vectorization. However the problem of contour extraction is complex and none of the existing schemes claim to be the best solution for the same. Now days the colour topographical maps are available and it is essential to recognize all its features. Recent papers deal scanned colour images using the mean and variance of the hue channel for discriminating soil types on a digitized soil map or transforming the input RGB colour space into another colour space taking the chromaticity into account. A majority of map analysis techniques have concentrated on binary maps that have thicker features which are spaced far apart from each other. Only few focused on colour images and thin closely spaced linear features. Dupont et al. used a water shed divide algorithm in RGB space to assign a pure map colour to each pixel. This algorithm performs well for image scanned by high resolution and quality scanners but not for the image which contains alias and false colour. Wu et al. used a multi-layer neural network to extract characters and lines from colour map images. Their inputs include features that comprehend color intensities and gradients. This approach does not figure out color aliasing and false colors inherent in the topographic maps. Hedly and Yan developed a gradient threshold method to overcome the aliasing and false colour problems. The final thinned raster image often consists of broken lines. These lines when vectorized and exported to standard CAD software like AutoCAD cause huge problems to the GIS user, because they have to identify and reconnect all the open (hanging) contour lines. Researches in the field of GIS have looked at various methodologies to resolve this problem. Image processing for cartography: a Review Analytical and computer cartography were used in the image processing applications earlier, but researchers start using the recent developments in the new areas like computational geometry and neural network to solve image processing problems. Conventional approaches split the raster to vector conversion process in the following four main steps – Raster image pre-processing, Raster Image segmentation and contour extraction, Raster Image post processing and skeletonization and Raster to Vector conversion (Fig3). The outputs of the first three steps are in the form of raster image and only in the final step DXF file (vector format) is generated.  Fig 3 Overview of conventional vectorization methodology ALGORITHM FOR VECTORIZATION PROCESS The following are the main features proposed algorithm
 Fig.4 Methodology for the Vectorization Process |