Disclosed is a method for concealing a dark defect in an image sensor. The method includes the steps of: setting a window including five sequential pixels from pixel data sequentially outputted, the third pixel and the fifth pixels being identical in Bayer color with each other and the second and the fourth pixels being identical in Bayer color with each other; calculating an average value .alpha. of the first and the fifth pixels and a plurality of .alpha./n obtained by dividing the average value .alpha. by n; selecting one of the plurality of .alpha./n; comparing a threshold value .alpha.+.alpha./n calculated by adding the average value .alpha. to the selected .alpha./n with a value of the third pixel in size; and omitting a concealment of the dark defect when the threshold value .alpha.+.alpha./n is greater than the value of the third pixel.