Mehmet Çetin Says
Hi all!!There are(were) alot of people looking for median filter and sort algorithm so were I. At last I wrote aprogram sorting any lentgh any number of data and median filtering. 16 24 or 32 bits 5 7 9 11 ....bytes data.
Thanks to Lawrence Lille. This program is a higher version of his 8 bit sort/filter program. Thanks for the idea.
; ;this is a plug and play code ;not a plug and pray code!! ; ;Thanks to Lawrence Lille for the original ;program and the idea ; ;register assignments ; cblock 0x20 ; x y ; analog1h analog1l ; analog2h analog2l ; analog3h analog3l ; analog4h analog4l ; analog5h analog5l ; analog6h analog6l ; analog7h analog7l ; analog8h analog8l ; analog9h analog9l ; analog10h analog10l ; analog11h analog11l ; analog12h analog12l ; analog13h analog13l ; analog14h analog14l ; analog15h analog15l ; ;result regs ; resulth resultl ; ;regs used to compare the adjacent data ;in the array to compare and swap the places ;to sort them in ascending or descending order ; ahi amd alo ; bhi bmd blo ; a1hi a1lo ; b1hi b1lo ; c1hi c1lo ; d1lhi d1lo ; ;data pointer counter ; cntr1 ; endc ; indir equ 00h tmr0 equ 01h pcl equ 02h status equ 03h fsr equ 04h ;***************status register bits #define c status,0 #define dc status,1 #define z status,2 #define pd status,3 #define to status,4 #define rp0 status,5 #define rp1 status,6 #define irp status,7 ; list p=16c620 ; pic16c620 is the target processor org 0 ; goto main ; ;the pointer table must be in ;accordance with the nember of the data ;array!! ; gethi addwf pcl,f retlw analog1h retlw analog2h retlw analog3h retlw analog4h retlw analog5h retlw analog6h retlw analog7h retlw analog8h retlw analog9h retlw analog10h retlw analog11h retlw analog12h retlw analog13h retlw analog14h retlw analog15h ; ; ; getlo addwf pcl,f retlw analog1l retlw analog2l retlw analog3l retlw analog4l retlw analog5l retlw analog6l retlw analog7l retlw analog8l retlw analog9l retlw analog10l retlw analog11l retlw analog12l retlw analog13l retlw analog14l retlw analog15l ; ;place some data to test ;the code it really works!! ; readpot2 movlw d'251' movwf analog1l movlw h'ff' movwf analog1h ; movlw d'254' movwf analog2l movlw h'ff' movwf analog2h ; movlw d'10' movwf analog3l movlw h'ff' movwf analog3h ; movlw d'200' movwf analog4l movlw h'ff' movwf analog4h ; movlw d'50' movwf analog5l movlw h'ff' movwf analog5h ; movlw d'4' movwf analog6l movlw h'ff' movwf analog6h ; movlw d'255' movwf analog7l movlw h'ff' movwf analog7h ; movlw d'99' movwf analog8l movlw h'ff' movwf analog8h ; movlw d'12' movwf analog9l movlw h'ff' movwf analog9h ; movlw d'11' movwf analog10l movlw h'ff' movwf analog10h ; movlw d'13' movwf analog11l movlw h'ff' movwf analog11h ; movlw d'9' movwf analog12l movlw h'ff' movwf analog12h ; movlw d'8' movwf analog13l movlw h'ff' movwf analog13h ; movlw d'7' movwf analog14l movlw h'ff' movwf analog14h ; movlw d'6' movwf analog15l movlw h'ff' movwf analog15h ; retlw 0 ; end readpot2 ;compare two adjacent variables by 16 to 32 or more ;bit math save them in temp regs and swap them if necessary ; ; x a ; y b ; x-y ; a-b ; order ;x a ; a := a1 := analog[cntr1] movf cntr1,W call getlo movwf fsr movf indir,W movwf alo movwf a1lo movf cntr1,W call gethi movwf fsr movf indir,W movwf amd movwf a1hi clrf ahi ; ;y b ; b := b1 := analog[cntr1+1]; incf cntr1,f movf cntr1,W call getlo movwf fsr movf indir,W movwf blo movwf b1lo movf cntr1,W call gethi movwf fsr movf indir,W movwf bmd movwf b1hi clrf bhi decf cntr1,f ; ; b-a ; y-x ; difference := a-b = analog[cntr1] - analog[cntr1+1]; ; Perhaps it would be better to use the routine at ; http://massmind.org/techref/microchip/math/sub/16bb.htm. movf blo,W subwf alo,f ; movf bmd,W skpc incfsz bmd,W subwf amd,f ; movf bhi,W skpc incfsz bhi,W subwf ahi,f ; if( difference < 0 ) ; { ; swap them ; } btfss ahi,7 goto ba_ ;x>y ab_ retlw 0 ; ;swap a b ; ;y>x ba_ ;x ; analog[cntr1] = b1; movf cntr1,W call getlo movwf fsr movf b1lo,W movwf indir movf cntr1,W call gethi movwf fsr movf b1hi,W movwf indir ;y ; analog[cntr1+1] = a1; incf cntr1,f movf cntr1,W call getlo movwf fsr movf a1lo,W movwf indir movf cntr1,W call gethi movwf fsr movf a1hi,W movwf indir decf cntr1,f retlw 0 ; end order() ;if there are n values ; ;x and y values are n-1 ; ; for( y=n-1; y>0; y-- ){ ; cntr1 = 0; ; for( x=n-1; x>0; x--){ ; order( analog[cntr1], analog[cntr1+1] ); ; cntr1++; ; } ; } sortarray movlw .14 ;.14 movwf y ; for y = 15 do while y >0 sortloop2 clrf cntr1 movlw .14 ;.14 movwf x sortloop call order ; sort fsr and fsr+1 incf cntr1,f ; increment fsr pointer decfsz x,f ; end do loop goto sortloop decfsz y,f goto sortloop2 ; ;the result is in the regs in the mid ;pos. of the data array that can be any length ;if n=15 median index is 8 ;your result is here!! ; movf analog8h,W ;here's your answer movwf resulth ;placed into the file result for use anywhere else. movf analog8l,W movwf resultl ; return ; end sortarray ;***************************************************************************** ;place the values and do the task ; main call readpot2 call sortarray ;sort the new data into the array lp1 goto lp1 ;that's all ; end
David A Cary Says: "
(I modified to use standard notation:
Use ",W" instead of ",0".
Use ",f" instead of ",1". ) "
David A Cary Says:
Rather than comparing *adjacent* pointers (doing a complete bubblesort), would the extra complexity of a (single iteration of) quicksort make it go any faster ? Especially when I collect lots of data, and I only want the median, not a completely sorted list.pseudo-code for single iteration of quicksort:
- partial_sort:
- scan the left half of the array for any "out of place item". (Any item with a values more than or equal to the value at the midpoint of the array). (if there is none, leave left_pointer pointing at the midpoint).
- scan the right half of the array for any "out of place item". (Any item with a value less than or equal to the value at the midpoint of the array). (if there are none, leave right_pointer pointing at the midpoint).
- swap the 2 out of place items.
- While left_pointer and right_pointer are *different*, repeat partial_sort.
- Otherwise, left_pointer and right_pointer are now both pointing at the midpoint of the array, which is the median. All the values to the left are less than the median (but are probably not sorted). All the values to the right are greater than the median (but are probably not sorted).
variations: rather than picking *any* out-of-place item, would it go any quicker to pick the *worst* out-of-place item (the maximum value on the left side, the minimum value on the right side) ?
I had great trouble with unstable adc values and this code helped a great deal, but it was not the final answer. I finaly averaged 16 median values to get a very stable result. With a 4MHz PIC it takes quite some time, but the result was worth it.+
For a practical application of this code see: PIC Microcontoller Input / Ouput Method - Median Filtering A2D results by Louis Nel
Questions:
file: /Techref/microchip/math/medfilsort-mc.htm, 8KB, , updated: 2006/10/31 17:05, local time: 2024/10/13 00:03,
owner: DAV-MP-E62a,
35.171.164.77:LOG IN ©2024 PLEASE DON'T RIP! THIS SITE CLOSES OCT 28, 2024 SO LONG AND THANKS FOR ALL THE FISH!
|
©2024 These pages are served without commercial sponsorship. (No popup ads, etc...).Bandwidth abuse increases hosting cost forcing sponsorship or shutdown. This server aggressively defends against automated copying for any reason including offline viewing, duplication, etc... Please respect this requirement and DO NOT RIP THIS SITE. Questions? <A HREF="http://www.piclist.com/techref/microchip/math/medfilsort-mc.htm"> PIC Microcontoller DSP Math Method median filter and sort </A> |
Did you find what you needed? |
PICList 2024 contributors:
o List host: MIT, Site host massmind.org, Top posters @none found - Page Editors: James Newton, David Cary, and YOU! * Roman Black of Black Robotics donates from sales of Linistep stepper controller kits. * Ashley Roll of Digital Nemesis donates from sales of RCL-1 RS232 to TTL converters. * Monthly Subscribers: Gregg Rew. on-going support is MOST appreciated! * Contributors: Richard Seriani, Sr. |
Welcome to www.piclist.com! |
.