modify pdf_to_img

@@ -32,3 +32,4 @@ data/*
 
 test*
 flow_test.py
+pdf_test.py
\ No newline at end of file

@@ -12,8 +12,10 @@ from unicodedata import normalize
 # 页面保存为png图片参数
 ZOOM_X_1 = ZOOM_Y_1 = 1.0
 ZOOM_X_2 = ZOOM_Y_2 = 2.0
+ZOOM_X_3 = ZOOM_Y_3 = 3.0
 trans_1 = fitz.Matrix(ZOOM_X_1, ZOOM_X_1).preRotate(0)  # zoom factor 1 in each dimension
 trans_2 = fitz.Matrix(ZOOM_X_2, ZOOM_X_2).preRotate(0)  # zoom factor 2 in each dimension
+trans_3 = fitz.Matrix(ZOOM_X_3, ZOOM_X_3).preRotate(0)  # zoom factor 3 in each dimension
 
 # 特殊filter处理
 ADOBE_FILTER_SET = {'FlateDecode', 'JPXDecode', 'JBIG2Decode'}
@@ -25,6 +27,12 @@ WH_COUPLE_3 = (100, 100)
 WH_COUPLE_4 = (100, 300)
 WH_COUPLE_5 = (100, 200)
 
+# 碎图宽度阈值
+TINY_IMG_MAX_WIDTH = 1800
+
+# 大图宽高阈值
+WH_COUPLE_6 = (1800, 1400)
+WH_COUPLE_7 = (2500, 3000)
 
 class PDFBuild:
 
@@ -165,8 +173,10 @@ class PDFHandler:
         except Exception as e:
             pass
 
-    def page_to_png(self, page):
-        if page.MediaBoxSize.x > 1500 or page.MediaBoxSize.y > 1500:
+    def page_to_png(self, page, is_big_img=False):
+        if is_big_img:
+            pm = page.getPixmap(matrix=trans_3, alpha=False)
+        elif page.MediaBoxSize.x > 1500 or page.MediaBoxSize.y > 1500:
             pm = page.getPixmap(matrix=trans_1, alpha=False)
         else:
             pm = page.getPixmap(matrix=trans_2, alpha=False)
@@ -247,6 +257,9 @@ class PDFHandler:
             if il[i][-1] in ADOBE_FILTER_SET:
                 page_to_png = True
                 break
+            if il[i][2] >= TINY_IMG_MAX_WIDTH:
+                page_to_png = True
+                break
         else:
             for i in range(length):
                 # 当图片对象够大时，不作碎图合并处理，而是单纯提取
@@ -446,6 +459,10 @@ class PDFHandler:
                             page = pdf.loadPage(pno)
                             self.page_to_png(page)
                         # 大图
+                        elif width >= WH_COUPLE_6[0] or height >= WH_COUPLE_6[1]:
+                            is_big_img = (width < WH_COUPLE_7[0] and height < WH_COUPLE_7[1]) # 防止图片过大
+                            page = pdf.loadPage(pno)
+                            self.page_to_png(page, is_big_img=is_big_img) 
                         elif xref not in self.xref_set:
                             self.extract_single_image(pdf, xref, smask, colorspace, pno)
                     # 3.页面图片对象数目大于1时，特殊处理