According to an embodiment, an image processing apparatus selects as an output image a candidate character component, from which a non-character component is removed, in a gradation having the largest number of pixels when there is a significant difference between the number of character pixels in the gradation having the largest number of character pixels and the number of character pixels in a gradation having the second largest number of character pixels, and selects as an output image a candidate character component, from which the non-character component is removed, in a gradation having the smallest number of edge pixels when there is no significant difference between the number of character pixels in the gradation having the largest number of character pixels and the number of character pixels in the gradation having the second largest number of character pixels.