Fruit shape is one of the most important quality parameters for evaluation by consumer performance. Also misshapen fruits are usually rejected according to grading standards of fruit. This case is more considered for watermelons, because of large size and sensitive skin. This study was carried out to determine “detection algorithm” for misshapen watermelons. Physical characteristics of watermelon such as mass, volume, dimensions, density, spherical coefficient and geometric mean diameter were measured. Relations and correlations coefficient obtained between above characteristics for normal and non-standard fruit shape. It was found that weight of normal watermelon could be determined by image analysis with error 2.42%. In addition fruit shape of long type watermelon (Charleston Gray cv.) in front view was well-described by an ellipsoid model with R2 = 0.97. Finally, the results indicated that length to width ratio and fruit area (2D) to background area ratio can be used to determine misshapen fruit.