Compressing images using contours

Abstract

A method for the vectorisation of digital images into contour maps with subsequent conversion of the contours to a pixel format is presented. This method may offer an alternative spatial image compression which is computationally inexpensive and can be directly applied to compressing certain classes of grey-scale and/or colour (RGB, 24 bits) imagery of any size. The feasibility of this study is based on research which shows that pixel based imagery can be sufficiently and accurately represented by their contour maps if a suitable contour model and scale selection method is used (Elder and Goldberg, 2001). The compression process is based on filtering and eliminating those contours that may contain redundant information. Contours extracted from digital images may contain multiple redundant data (i.e. intersecting or nested contours), any of which might logically be used as a basis for discrimination or, on the other hand, used in the reconstruction of an original captured image. As per current image compression techniques the proposed method does not require special hardware and, if combined with existing encoding schemes, it can be efficiently used for image transmission purposes due to its relatively modest storage requirements. Depending on the applications, and the amount of contour lines employed in the reconstruction of an image, the process allows for various levels of image accuracy while preserving visual integrity and reducing compression artifacts