Background interferences due to text-like objects make the text binarisation process difficult. This paper presents a method for text binarisation which works across the localised text region from heterogeneous colour images with multi coloured texts and complex background with text-like objects. The method also supplements the variation in font size and uneven illumination of text. In the proposed integration of edge and colour analysis (IECA) algorithm, edge and colour information are integrated to binarise the text. Background interferences due to textured background are reduced with the character size uniformity check (CUC) algorithm and the edge quadrant coverage analysis (EQCA) algorithm. Then some background patches within the segmented character box are removed and image is binarised by corner vertices colour analysis (CVCA) algorithm. The proposed method has been applied on localised region from scene and caption text images and compared with Otsu, Niblack and Kasar method and shown encouraging performance of the proposed method. It is also shown that IECA algorithm can successfully binarise the images with non-uniform illumination, complex/textured background, multi coloured texts and existence of text-like objects. The proposed IECA approach is the required preprocessing for optical character recognition (OCR) and multi media processing.