Re: identifying yoghurt jars


Reija wrote:
As to color matching - if you don't want to pay the price of 3D color
matching you may want to consider using H and S (or similar metrics) while
eliminating pixels with very low S and I values.

So just to convert from RGB to HSV, then eliminate pixels with low S
and V(or I) and then to calculate histogram for remaining pixels using
only H and S? That would make calculating lighter, thanks for the hint!

But with shape detection I don't understand yet how to compare two
shapes. If I find the edges with Canny or something like that, isn't
the output just an image of the edges? So how do I measure the
similarity between to shapes? What would be good search words for
Google?


feature extraction, shape, shift invariant, scale invariant, rotation
invariant.

There are many good textbooks on image processing; others on pattern
recognition. As regards image processing equivalents of Numerical
Recipes, there is Myler and Weekes, The Pocket Handbook of Image
Processing Algorithms in C. Also books by J.R. (Jim) Parker and Timothy
Masters.

But try to gain a good foothold in image processing first. The fact
that you know about Numerical Recipes indicates that this should be
easy for you. Books like, e.g., Castleman, Gonzalez and Woods.
http://www.cee.hw.ac.uk/hipr/html/welcome.html is good.

Try to separate the process into steps like:

- preprocessing;

- segmentation; I'd tend to attempt grey-level based segmentation
first, before trying an edge based approach;

- feature extraction;

- classification.

Maybe some help in:

http://www.jgcampbell.com/ip/pr.pdf

I know that ad hoc procedures /can/ work, but you need to be very
clever and have very good intuition for the area.

Incidentally, another poster hints at avoiding colour if possible; I
agree; all sorts of problems with calibration -- but maybe I'm showing
my age.

Best regards,

Jon C.

.

Relevant Pages

  • Re: location of segmented image
    ... I am sorry for this image processing newbie question. ... Based Segmentation Using K-Means Clustering" demo of image processing ... That should give you the logical indexes of the "blue" pixels. ...
    (comp.soft-sys.matlab)
  • The use of PCA on a 64x64 Matrix
    ... I'm new to image processing, and am having trouble fitting my head ... value of the pixels. ... and finding the mean for each column vector. ... I then calculated the eigenvectors ...
    (comp.graphics.algorithms)
  • Re: Desperately needing direction with saving images under different compression/quality schemes
    ... KB's and open it in ms paint, then save it without making any changes ... further changes to pixels, maybe just a few bytes of changes in headers ... but thats it). ... background is not in image processing, i am not sure what these changes ...
    (sci.image.processing)
  • Re: Specific area inside of an image
    ... SO what this all boils down to is that I need to find an area ... inside of this circle that has 16 pixels of the same value touching ... Well, conceptually doesn't seem that difficult--what there is in later versions of ML or toolboxes associated w/ image processing that might simplify the coding I've no clue as image processing isn't my bag, but... ...
    (comp.soft-sys.matlab)
  • Re: Determining row of pixels among a group that has the max number of pixels above a threshold
    ... It consists of 30 groups of 1x10 pixels. ... I know image processing software but not the hardware. ... who can design you a custom ASIC chip. ...
    (sci.image.processing)