Re: umax-backend

Rogier Wolff (R.E.Wolff@BitWizard.nl)
Mon, 1 Sep 1997 22:37:36 +0200 (MET DST)

Jonathan A. Buzzard wrote:
>
>
> R.E.Wolff@BitWizard.nl said:
> > Yes, the q/a program will show "only noise" still, the doctors won't
> > allow you to trhow away that information. Maybe if you average 4
> > pixels together that you get a little information. This IS visible to
> > the naked eye.....
>
> Never said the radiologist would let you throw the info away, but noise
> is noise and averaging bits together will (should)

should not make a difference. Right.

> > How do you measure the "just noise" in a 12bit scanner? I'd try to
> > generate a perfect analog greyscale and scan that.
>
> Not that I was involved in the q/a program, (it was an M.Sc. project done
> at the same time as mime, and I had to use the same scanner for some
> radiotherapy films). Roughly their was a test film which had a large number
> of squares of different (but uniform) optical density. Roughly a scan was
> taken of the film, a template was placed over the image and some statistics
> where done on the pixels of the different areas. As far as I understood
> the analysis showed that on a uniform area the distribution of the values
> in the lower two bits was more or less uniform. The conclusions drawn from
> that it must be noise.

If "that is all there is to it", then I don't think this is a valid
conclusion. I think I'd have to read the whole thesis.

My explanation of the "uniformity" of the values in the lower bits is
"there is noise in the source image".

Yeah, it may look VERY uniform, but we're trying to detect the noise
in the area that we're agreeing on that humans (almost) don't see the
difference any more.

Try creating a greyscale image: example:
position -> 1234556789
/ 1234556789
_/ 1234556789
^ / 1234556789
intensity| / 1234556789

and notice the white stripe.

% xv -perfect -geometry 260x240 -
P2
260 1 255
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26
27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49
50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72
73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95
96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113
114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 128 128
128 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143
144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160
161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177
178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194
195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211
212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228
229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245
246 247 248 249 250 251 252 253 254 255
^D (EOF)

The only thing I changed from the ramp is that the value 128 is
mentioned 5 times. The human eye makes that into a WHITE band.

(This effect requires 8 bit deep DAC's.)

What might be a reasonable test is to scan the same image over and
over again. For every position in the image you can then calculate
the statistical parameters for that position. The standard deviation
has something to do with the amount of noise. I expect the pixels
on sharp edges to show a higher noise figure than other pixels
due to mechanical positioning problems.

Roger.

--
Source code, list archive, and docs: http://www.azstarnet.com/~axplinux/sane/
To unsubscribe: mail -s unsubscribe sane-devel-request@listserv.azstarnet.com