Colour reduction of digital images in its common uses is used for reducing image processing time or for the possibility of printing images of appropriate quality by a printer; in these applications the overall image quality is considered. In some other applications colour reduction is merely done to choose the suitable colour and this is the colour accuracy of the chosen colours which is significant; such as choosing colour for military camouflage uniforms to have better colour similarity with its surrounding or hiding out a structure in urban environment to improve the beauty of a city. In such applications the most important factor is the accuracy of the colour selection. In the present research four images of different forest environments were chosen and to select camouflage colours, K-means clustering, fuzzy C-means clustering and diversity methods were performed. The K-means clustering method was performed in CIEL * a * b * colour space by applying CIELAB and CIEDE2000 colour difference formulas and it was also performed in RGB colour space by Euclidean distance formula. It is noteworthy to mention that the RGB values of the pixels of the images firstly transformed to CIEL * a * b * values for the accuracy of color reproduction both in the digital camera and in the LCD monitor used in the present research. The LCD monitor was characterized using a Lacie Blue eye pro colourimeter and its software package. Then the forest images were printed with high quality, and photographed in a calibrated condition, and the colours of the images were reduced into 4 colours using the aforementioned colour reduction methods. By using a Matlab GUI environment, each of the photographed forest images was considered as the background of the main image of the experiment and on each of the backgrounds 35 different locations were chosen. A small camouflaged image was placed on each of the 35 locatio the small camouflaged images had an Iranian forest military pattern whose colours were selected by the colour reduction methods, done on the 4 chosen forest images. The observers, whose visions were examined by an Ishihara test, were asked to sit in a darkened room in front of the LCD. After the adaptation of their eyes to the darkness, they were shown the images. The observers were asked to click on the small camouflaged images as soon as they manage to detect it and the time of the detection was recorded by the program. The results of this visual test signifies an increase in the detection time as well as the percentage of failure to detect camouflage patterns with the K-means clustering technique in CIEL * a * b * ace and CIEDE2000 colour difference formula in comparison to the other methods. Generally, the K-means clustering method in CIEL * a * b * colour space with CIELAB colour difference formula and the fuzzy method have the second and third rank respectively. Whereas, the K-means clustering method in RGB colour space was ranked fourth and the worst results belong to the diversity method. Key Words: Camouflage, Digital Image, Clustering, Colour space, Colour Reduction