|
上次讲到了用DIB方法来获取图像的像素。从这次开始将如果运用已经得到的像素来处理图像。
图像插值放大的方法有很多，最主要的有二次线性插值和三次线性插值这两种。
这次我把自己的程序中所用的二次线性插值的算法公布给大家，希望对各位要使用VB写类似程序的朋友有所帮助。
程序中用到的API、数据类型、全局变量的定义请参考上一篇：
《VB图像处理，（一）像素的获取和输出 》
Public Sub ZoomImage(ByVal OutPutWidth As Long, ByVal OutputHeight As Long)
Dim I As Long
Dim L As Long
Dim X As Long
Dim Y As Long
Dim Xb As Long
Dim Yb As Long
Dim Xe As Long
Dim Ye As Long
Dim M As Integer
Dim N As Integer
Dim CurR As Long
Dim CurG As Long
Dim CurB As Long
Dim NxtR As Integer
Dim NxtG As Integer
Dim NxtB As Integer
Dim DR As Single
Dim DG As Single
Dim DB As Single
Dim DRt As Single
Dim DGt As Single
Dim DBt As Single
Dim Xratio As Single
Dim Yratio As Single
Dim CurStep As Single
Dim NxtStep As Single
Dim NegN As Single
On Error GoTo ErrLine
If Not CanZoom Then Exit Sub
Done = False
OutPutWid = OutPutWidth - 1
OutPutHei = OutputHeight - 1
I = (Bits \ 8) - 1
ReDim ColTmp(I, InPutWid, OutPutHei) ''''先从Y方向进行缩放处理，结果保存在此中间数组内
ReDim ColOut(I, OutPutWid, OutPutHei)
Xratio = OutPutWid / InPutWid
Yratio = OutPutHei / InPutHei
TimeZoom = timeGetTime
NegN = 1 / Int(Yratio + 1)
For X = 0 To InPutWid
CurR = ColVal(0, X, 0)
CurG = ColVal(1, X, 0)
CurB = ColVal(2, X, 0)
CurStep = 0
NxtStep = 0
For Y = 0 To InPutHei - 1
NxtStep = CurStep + Yratio
Yb = CurStep
Ye = NxtStep
N = Ye - Yb
ColTmp(0, X, Yb) = CurR
ColTmp(1, X, Yb) = CurG
ColTmp(2, X, Yb) = CurB
M = Y + 1
NxtR = ColVal(0, X, M)
NxtG = ColVal(1, X, M)
NxtB = ColVal(2, X, M)
If N > 1 Then
DRt = (NxtR - CurR) * NegN
DGt = (NxtG - CurG) * NegN
DBt = (NxtB - CurB) * NegN
DR = 0
DG = 0
DB = 0
For L = Yb + 1 To Ye - 1
DR = DR + DRt
DG = DG + DGt
DB = DB + DBt
ColTmp(0, X, L) = CurR + DR
ColTmp(1, X, L) = CurG + DG
ColTmp(2, X, L) = CurB + DB
Next
End If
CurStep = NxtStep
CurR = NxtR
CurG = NxtG
CurB = NxtB
Next
ColTmp(0, X, OutPutHei) = NxtR
ColTmp(1, X, OutPutHei) = NxtG
ColTmp(2, X, OutPutHei) = NxtB
Next
NegN = 1 / Int(Xratio + 1)
For Y = 0 To OutPutHei
CurR = ColTmp(0, 0, Y)
CurG = ColTmp(1, 0, Y)
CurB = ColTmp(2, 0, Y)
CurStep = 0
NxtStep = 0
For X = 0 To InPutWid - 1
NxtStep = CurStep + Xratio
Xb = CurStep
Xe = NxtStep
N = Xe - Xb
ColOut(0, Xb, Y) = CurR
ColOut(1, Xb, Y) = CurG
ColOut(2, Xb, Y) = CurB
M = X + 1
NxtR = ColTmp(0, M, Y)
NxtG = ColTmp(1, M, Y)
NxtB = ColTmp(2, M, Y)
If N > 1 Then
DRt = (NxtR - CurR) * NegN
DGt = (NxtG - CurG) * NegN
DBt = (NxtB - CurB) * NegN
DR = 0
DG = 0
DB = 0
For L = Xb + 1 To Xe - 1
DR = DR + DRt
DG = DG + DGt
DB = DB + DBt
ColOut(0, L, Y) = CurR + DR
ColOut(1, L, Y) = CurG + DG
ColOut(2, L, Y) = CurB + DB
Next
End If
CurStep = NxtStep
CurR = NxtR
CurG = NxtG
CurB = NxtB
Next
ColOut(0, OutPutWid, Y) = NxtR
ColOut(1, OutPutWid, Y) = NxtG
ColOut(2, OutPutWid, Y) = NxtB
Next
Done = True
TimeZoom = timeGetTime - TimeZoom
CanPut = True
Exit Sub
ErrLine:
MsgBox Err.Description
End Sub
全局变量定义：
Dim ColTmp() As Byte ''''用于保存插值中间变量
Dim OutPutHei As Long ''''要插值的目标高度
Dim OutPutWid As Long ''''要插值的目标宽度
Public TimeZoom As Long ''''插值运算使用的时间
简单解释一下关于二次线性插值算法。
（为了说明算法本身，我们只计算这个图片的红色分量，因为红绿蓝三种颜色的计算方法完全相同）
假设我们有一个很简单的图片，图片只有4个像素（2*2）
A B
C D
现在我们要把这个图片插值到9个像素：3*3
A ab B
ac abcd bd
C cd D
其中大写的字母代表原来的像素，小写字母代表插值得到的新像素。
想必看到这个图，大家心里已经有了这个算法了。
ab=(A+B) / 2
cd=(C+D) / 2
ac=(A+C) / 2
bd=(B+D) / 2
abcd=(ab+cd) / 2＝(A+B+C+D) / 4
推导：ab= A + (B-A) / 2
cd=C +(D-C) / 2
...
很简单，对吧，先从一个方向把只涉及两个原始像素的新像素算出来。我们这里假定先计算水平方向。
而在算垂直方向的插值的时候，因为ab和cd已经在前面算好了，所以abcd的计算也和计算ac和bd没有任何区别了。
有可能为有朋友已经想到把原来的图像插值到4*4或5*5的方法了。
A ab1 ab2 B
ac1 ab1cd11 ab2cd21 bd1
ac2 ab1cd12 ab2cd22 bd2
C cd1 cd2 D
推导：ab1 = A + (B-A) * 1 / 3
ab2 = A + (B-A) * 2 / 3 =ab1+(B-A) / 3
cd1 = C + (D-C) * 1 / 3
cd1 = C + (D-C) * 2 / 3 =cd1+(D-C) / 3
...
以A和B为例，先求出原始像素的差(A-B)再算出每一步的递增量(A-B) / 3
然后每一个新的点就是在前面那个点的值加上这个递增量就是了。
这里我们假设A=100, B=255 放大倍率为3，水平方向插值
先计算出原始像素的差 [1] [2] 下一页
没有相关教程
|
注：本站部分文章源于互联网，版权归原作者所有！如有侵权，请原作者与本站联系，本站将立即删除！ 本站文章除特别注明外均可转载，但需注明出处！ [MinTao学以致用网] 
