package colors type LightMap []Luminance // Hex returns all luminance value as a hex encoded string. func (m LightMap) Hex() (result string) { for _, luminance := range m { result += luminance.Hex() } return result } type diffValue struct { a []int b []int } // Pixel comparisons for calculating the diff value. var diffValues = []diffValue{ {a: []int{4, 4, 4, 4}, b: []int{1, 3, 5, 7}}, {a: []int{0}, b: []int{1}}, {a: []int{0}, b: []int{3}}, {a: []int{2}, b: []int{1}}, {a: []int{2}, b: []int{5}}, {a: []int{6}, b: []int{3}}, {a: []int{6}, b: []int{7}}, {a: []int{8}, b: []int{5}}, {a: []int{8}, b: []int{7}}, } /* Alternative values to experiment with: {a: []int{4, 4, 4, 4}, b: []int{1, 3, 5, 7}}, {a: []int{0}, b: []int{2}}, {a: []int{8}, b: []int{6}}, {a: []int{0}, b: []int{1}}, {a: []int{1}, b: []int{2}}, {a: []int{2}, b: []int{5}}, {a: []int{8}, b: []int{7}}, {a: []int{7}, b: []int{6}}, {a: []int{5}, b: []int{8}}, {a: []int{6}, b: []int{3}}, Other ideas: - Use Lightness instead of luminance - Use more pixels (difficult as we only have 3x3 thumbs right now) - Iterative calculation using smaller offsets for each round (current offset is +4) For a more complex implementation, see https://github.com/EdjoLabs/image-match/blob/master/image_match/goldberg.py http://www.cs.cmu.edu/~hcwong/Pdfs/icip02.ps General blog post on Perceptual image hashing: https://jenssegers.com/perceptual-image-hashes */ // Diff returns an integer that can be used to find similar images. func (m LightMap) Diff() (result uint32) { if len(m) != 9 { return 0 } result = 1 for _, val := range diffValues { result = result << 1 a := 0 b := 0 for _, i := range val.a { a += int(m[i]) } for _, i := range val.b { b += int(m[i]) } if a+4 > b { result += 1 } } return result }