Chromatic blur perception

1. Chromatic blur perception Frederick A. A. Kingdom McGill Vision Research, McGill University, Montréal

2. Q. Does colourvision contribute to the perception of blur?

3. Blur is important for depth perception

4. Tilt-shift photography

8. ‘Blue-yellow’ (BY) ‘Red-green’ (RG)

9. L+M L-M S-(L+M) LUM ‘RG’ ‘BY’

10. Blur discrimination thresholds * = s Wuerger, Owens & Westland (2001) JOSA A, 18, 1231-1239

11. Blur discrimination thresholds s s + Ds Wuerger et al. (2001) JOSA A, 18, 1231-1239

12. s Ds LUM RG BY Wuerger et al. (2001) JOSA A, 18, 1231-1239

13. LUM RG low contrast medium contrast high contrast

14. Results… Threshold Ds = < LUM RG BY Wuerger et al. (2001) JOSA A, 18, 1231-1239

15. Blurring an image of an everyday scene

16. Original RG blurred BY blurred LUM blurred Wandell (1995) Foundations of vision

17. Raw image Chromatic layer Luminance layer

19. Chromatic blurred Luminance blurred

20. Some possibilities……… 1. The luminance pattern masks the perceived blur in the chromatic pattern 2. We do not have a “sense” of chromatic blur, even though we can discriminate chromatic blur 3. Natural scenes versus laboratory stimuli

21. Gabor texture

22. no blur medium blur high blur

23. LUM RG BY

24. Blur level 0 1 2 3 LUM RG BY

25. LUM blur 0 1 2 3 0 1 BY blur 2 3

26. Methodology • Variation on paired comparisons • Appearance not performance task/measure • Both chromatic and luminance layers present in stimulus

27. “One different” Only one of BY or LUM different BY 2 LUM 1 BY 3 LUM 1

28. “Both different” Both BY and LUM different BY 0 LUM 2 BY 1 LUM 0

29. Results for EC RG only LUM only

30. Results for EC RG different RG different, LUM same

31. Results for EC LUM different, RG same “One different” RG different, LUM same

32. Results for EC “Both different” RG different, LUM different

33. Results for EC “One different” “Both different”

34. Results for EC BY only LUM only

35. Results for EC BY different BY different, LUM same

36. Results for EC “One different” “Both different”

37. Average slopes across 4 subjects: RG v LUM

38. Average slopes across 4 subjects: BY v LUM

39. Summary • Previous studies show that we are good at discriminating • the blur of LUM and isoluminant RG patterns, but not so good at • discriminating blur in isoluminant BY patterns. • In images of everyday scenes, we experience a strong sense of • blur when the luminance, but not chromatic layers are blurred. • Using Gabor textures with different combinations of chromatic • and luminance blur, subjects based their judgements of relative • blur on the LUM not RG layer, even though the different RG • blurs were discriminable i.e. not masked by the LUM layer.

40. L-M M-L Response Stimulus size Mullen & Losada (1999). Vis. Res., 39, 721-731

41. Georgeson et al. (2007), JOV, 7(13):7,1-21.

43. Conclusions • Even though we are able to discriminate different levels of • chromatic blur, our “sense” of chromatic blur is very weak or • non-existent. • Put another way…… The function relating perceived blur to • physical blur is steep for luminance but near-flat for chromatic stimuli. • Lack of a sense of chromatic blur may be because chromatic blur • Is redundant.

44. Sinusoidal scale distortion

46. Chromatic distorted Luminance distorted

47. Overall conclusion To understand how colour vision contributes to the perception of image structure, one needs to combine performance measures obtained at isoluminance with appearance measures using combined chromatic + luminance stimuli.

48. Acknowledgements Jason Bell, Faculty, Australian National University Alysha Bartsch, Undergraduate, McGill Christian Haddad, Medical student, McGill

49. Example forced-choice pair BY 3 LUM 0 BY 0 LUM 3

50. Example RG v LUM forced-choice pair